DOI: 10.18195/issn.0313-122x.78(1).2010.205-246 Records of the Western Australian Museum, Supplement 78: 205–246 (2010). An arid zone awash with diversity: patterns in the distribution of aquatic invertebrates in the Pilbara region of Western Australia

Adrian M. Pinder1, Stuart A. Halse1, 2, Russell J. Shiel3 and Jane M. McRae1, 2

1Department of Environment and Conservation, PO Box 51, Wanneroo, Western Australia 6946, Australia

2Present address: Bennelongia Pty Ltd, 5 Bishop St, Jolimont, Western Australia 6014, Australia

3School of Earth and Environmental Sciences, University of Adelaide, Adelaide, South Australia 5005, Australia

Corresponding author: email: [email protected]

Abstract – The Pilbara region of Western Australia is an arid zone with an abundance of wetlands, ranging from springs and river pools to salt marshes, claypans, rockpools and gnammas. The Pilbara is also rich in mineral resources, which are being heavily exploited, and most of the region is under pastoral lease, with both land uses potentially having an impact upon aquatic systems. In order to provide information for improved conservation planning, including environmental impact assessment and reserve system design, a survey of wetland fauna was undertaken as part of a broader biodiversity survey of the region. This paper describes patterns in the distribution of the region’s aquatic invertebrate diversity and identifi es environmental correlates of those patterns. Invertebrates and environmental variables were surveyed at 100 wetlands in spring and autumn between 2003 and 2006. The Pilbara appears to have a particularly diverse fauna for an arid zone, with just over 1000 species collected, an average of 94 species per sample and a maximum sample richness of 226 species. About a fi fth of the fauna is known to date only from the Pilbara. The remaining species are western endemics, occur elsewhere in northern and/or inland Australia, or have continental or wider distributions. Most species seem to be widespread in the Pilbara region but rare and/or restricted elements are present, especially in some permanently fl owing springs, including those in Millstream and Karijini National Parks, and in ephemeral wetlands such as Fortescue Marsh and freshwater claypans. Faunal composition was not strongly correlated with season, drainage divides, stream order or previously recognised bioregional boundaries. A number of assemblages of species with common patterns of occurrence were identifi ed, each showing unique associations with measured environmental variables. Flow, turbidity, salinity, sediments, macrophytes and hydrological persistence are among the environmental gradients most strongly correlated with occurrence patterns in the fauna.

INTRODUCTION is currently hampered by very poor knowledge of how aquatic biodiversity is distributed across The Pilbara region of north-western Australia the region (McKenzie et al. 2002). While a few of is a geologically and geomorphically distinct and the more iconic wetland systems are encompassed ancient part of the continent (Taylor 1994). It has by existing reserves (gorge streams in Karijini a harsh climate, with low but mostly seasonal National Park and the springs and deep permanent rainfall, resulting from intense tropical lows river pools in Millstream National Park), the in summer (many of cyclonic intensity). The conservation estate covers just 7.75% of the Pilbara combination of these elements, plus extensive groundwater aquifers with surface expressions, IBRA region in four parks and is not representative has created a diverse array of wetlands, including of the range of wetlands present in the region. claypans and clay fl ats, rockpools in creeks and This survey of aquatic invertebrates was part of rocky outcrops, springs, river pools (many of a broader project (McKenzie et al. 2009), designed which are permanent) and the large endorheic to provide information on the distribution of the Fortescue Marsh system. Land use in the Pilbara is region’s biodiversity for a range of conservation almost entirely pastoralism and mining and these management needs, including reserve system activities place considerable pressure on the region’s design and environmental impact assessment. wetlands. Effective protection of these wetlands Protection of biodiversity is achieved more easily 206 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 1 Maps of the Pilbara region showing wetland locations, major rivers, IBRA sub-regions (shaded, with full names) and drainage basins (bounded by bold lines and abbreviated as follows: ASH, Ashburton; OC, On- slow Coast; FOR, Fortescue; PHC, Port Hedland Coast; DGO, De Grey-Oakover; and GSD, Great Sandy Des- ert. Numbers next to sites in the bottom map are site numbers as per Table 1. Distribution patterns of aquatic invertebrates 207 and effi ciently where there is an understanding of tributaries (Figure 1). This area of about 225,000 km2 what it comprises and how it is distributed across includes all of the Pilbara IBRA bioregion as well the landscape (Haila and Margules 1996; Vinson as the some of the northern Gascoyne IBRA region and Hawkins 1998; Gaston 2000; Margules et al. (parts of the Augustus and Ashburton subregions) 2002). It is assisted further by an understanding (Australian Government 2004). A summary of the of how biodiversity patterning is related to region’s landforms, geology and climate is provided environmental gradients to allow extrapolation by McKenzie et al. (2009). beyond the limited area sampled (Nicholls 1989). The Pilbara region coincides largely with the This type of information is essential to designing Pilbara Craton, one of Australia’s major geological an effi cient conservation reserve system (Brooks blocks (Geological Survey of Western Australia et al. 2004; Pressey 2004), undertaking accurate 1990). The Craton is characterised by hard rock threatened species and communities assessments landscapes that were laid down in Archaean times, (Strayer 2006) and enabling the environmental 2.5 to 3.5 billion years ago, making them some of impacts of land management decisions to be the oldest rocks on the planet (Boulter 1986). The assessed in a regional context (Stork et al. 1996). basement is exposed as granite and greenstone Such work also provides information on ecological terrane in the north but overlain by rugged tolerances, which have direct relevance for sedimentary strata, volcanic fl ows and lateritised conservation by providing explanation of changes caps in the south. in animal abundance and distribution (e.g. Pinder Average annual rainfall is < 350 mm, falling et al. 2005) mostly in summer as a result of tropical lows and In Western Australia (WA), comprehensive cyclones that cross the Pilbara coast about every regional scale aquatic biodiversity assessments have second year on average (http://www.bom.gov.au). been carried out in the Carnarvon Basin, south of Average summer rainfall is 150–300 mm across the Pilbara (Halse et al. 2000; Keighery et al. 2000) most of the Pilbara but attenuates south-westward and in the South-west Agricultural Zone (Halse to < 100 mm. Winter rainfall, which occurs as et al. 2004; Lyons et al. 2004; Pinder et al. 2004). cold fronts pass over the region, has an opposing Biodiversity information from these surveys has trend with an average of 50–75 mm in the west contributed to conservation programs such as the and uplands to < 25 mm in the east and lowlands. selection of ‘biodiversity recovery catchments’ for Average maximum temperatures range from intensive management in the wheatbelt (Walshe 21–30ºC in winter to 36–42ºC in summer (Bureau of et al. 2004, Quinlan et al. 2009) and selection of Meteorology Australia 2008). land for the conservation estate in the Carnarvon In late summer 2002/3 (about six months prior Basin (McKenzie et al. 2000, Brandis 2009). The to sampling commencing in August 2003), several dataset presented here has already been used for low intensity cyclones and tropical lows caused conservation value assessment (e.g. Pinder and fl oods in the Ashburton River and upper (but not Leung 2009; Van Leeuwen 2009) and assessment lower) Fortescue River, and minor flooding in of likely environmental impacts of resource the De Grey-Oakover River. In early March 2004, development proposals in the Pilbara. Cyclone Monty caused severe fl oods in most major This paper examines the distribution of aquatic Pilbara river systems (Figure 2), except the Robe invertebrates across the region in relation to a River. In late March 2004, Cyclone Fay also caused range of landscape and environmental variables. severe fl oods, mostly in eastern rivers such as the The main approach taken is to identify groups De Grey and Coongan. Cyclones Clare and Glenda of species (assemblages) with similar patterns of in January and March 2006 resulted in severe distribution across space and time and presumably, fl oods in the Ashburton, Robe, Fortescue and Port therefore, similar responses to environmental Hedland coast Rivers but only minor fl ooding in gradients. The advantage of identifying groups of the De Grey-Oakover system. No cyclones affected species with similar distribution, before analysing the Pilbara in summer 2004/5. Winter rains in the influence of environmental variables on 2005 resulted in minor fl ow-pulses across most of occurrence, is that gradients and thresholds may the Pilbara (but these are not visible at the scale of be identifi ed for the subsets that are masked or Figure 2). underestimated when working with all species together (McKenzie et al. 2004). Wetland settings Many of the wetland systems in the Pilbara are The Pilbara region strongly infl uenced by the rapid movement of water For the purposes of this study, what we call in the landscape resulting from the combination the ‘Pilbara’ region is bounded to the south by of intense rainfall events (especially from cyclonic the main channel of the Ashburton River and to systems) and the largely impervious geology. the east by the De Grey-Oakover River and its In addition, the extensive alluvial, sedimentary 208 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae and fractured rock aquifers (Water and Rivers persist. Groundwater discharge at springs also Commission 1996; Johnson and Wright 2001) creates short sections of permanently flowing contribute large quantities of groundwater to creeks, sometimes within the broad channels of surface systems. Flash fl ooding of smaller, rocky middle-order rivers, where they support stands of headwater creeks, where fl ow lasts just hours to Melaleuca woodlands. days, often leaves small temporary to permanent The Pilbara has few large fl oodplains except on rockpools (the latter especially in gorges). Such the Roebourne Plains, where fl oodwaters from all events may occur in either summer or winter but rivers spread out, and in the middle and upper winter rainfall is usually insufficient to cause Fortescue Valley (downstream of Newman). Smaller signifi cant fl ows in the larger river channels. More fl oodplain areas include the upper catchment of substantial rain events, primarily in late summer, Duck Creek north-west of Karijini National Park result in catastrophic floods, with peak flows and parts of the lower Ashburton River upstream of 3 -1 frequently in the 100s to 1000s m .s in lowland the coastal plain. In fl oodplain areas, claypans and reaches of larger rivers. The hydrographs of these clay fl ats retain water after fl oods but some also scouring floods have very steep rising limbs receive water from feeder creeks and local runoff (lasting only days after a cyclone) with declining when broadscale fl ooding does not occur. Claypans fl ows that last for a few weeks, before the rivers are usually small, shallow and ephemeral, although recede to numerous temporary to permanent river the larger ones retain water for many months after pools. Rivers tend to be broad (up to a kilometre fl ooding. A regionally signifi cant and very large in width in lowland areas) and frequently have fl oodplain wetland is the Fortescue Marsh, which braided low-fl ow paths, mostly within the banks of is a terminal wetland for the upper Fortescue a single channel. Some smaller creeks, especially River, 100 km long, bounded downstream by the in the broader valleys and coastal plains, contain Goodiarie Hills. highly turbid billabongs and are herein referred to as ‘turbid pools’. Permanent pools usually Drainage basins occur where bedrock structures impede hyporheic There are fi ve drainage basins in the Pilbara groundwater flow, where springs discharge (Figure 1). The Ashburton River is 700 km long groundwater or where cliffs cause fl ow to scour and drains the southern slopes of the Hamersley pools that are sufficiently deep and shaded to Range and northern slopes of ranges to the south

Figure 2 Daily discharge for the period 2003 to 2006 for A, lower Ashburton River at Nanutarra; B, lower Fortescue at Gregory Gorge; and C, lower De Grey River at Coolenar Pool. Arrows above the graphs indicate timing of cyclones (M = Monty; F = Fay; C= Clare; G = Glenda). Distribution patterns of aquatic invertebrates 209 of the Pilbara, such as the Collier and Kenneth These major tributaries drain the eastern Chichester Ranges. For most of its length it consists of only one Ranges and numerous minor ranges within the main channel supplied by numerous, mostly short basin. The Oakover has some small tributaries (50–200 km) tributaries. The main channel contains arising from the Isabella and Gregory Ranges (site many large permanent pools (e.g. sites 49 and 53). 62) that form the eastern boundary to the Pilbara. The Cane and Robe Rivers, flowing off the Few off-channel wetlands occur in the basin, other western-most slopes of the Hamersley Range, than near the lower De Grey (site 35), so springs dominate the Onslow drainage basin, along with (site 63) and river pools (site 57) predominate. a few smaller channels that mostly contain water Finally, several pools and springs in river only in turbid pools (e.g. site 14) and do not reach channels (sites 27 to 29) in the Rudall River the coast as defi ned channels. The Robe catchment catchment were sampled. This system lies in the contains a number of springs in river channels such Great Sandy Desert Basin to the east of the Pilbara. as sites 15 and 16. The Fortescue River, 700 km long, has few METHODS significant tributaries and drains the southern slopes of the Chichester Range and the northern Site selection and far eastern slopes of the Hamersley Range. The upper Fortescue is essentially endorheic, In this study, a ‘site’ refers to a reach or segment terminating at the large, brackish to mesosaline of wetland about 200 m long. Most wetlands were Fortescue Marsh (sampled at sites 2 and 3, Figure small enough to be confi dent that the site sampled 1), but with fl oods also spreading out to the east included all of the major habitats present at the of the marsh and fi lling claypans and small creek wetland. A total of 100 sites in 98 wetlands were pools, such as sites 41 and 42. The marsh would selected to represent the diversity of wetlands have to be at least 13 m deep to overfl ow through the present within each drainage basin, with sites Goodiarie Hills downstream of the marsh and that distributed across all stream orders and along the has never been recorded. The Fortescue thus arises length of the major channels (Table 1, Figure 1). anew downstream of the marsh, where tributaries running off the slopes of the middle Hamersley and Sampling Chichester Ranges (e.g. site 94) fl ow into a valley A total of 189 samples were collected from the 100 containing several claypans (e.g. sites 4 and 40). This sites between spring 2003 and spring 2006 (Table middle section of the Fortescue is poorly defi ned, 1). Most sites were sampled once in autumn (late with a braided channel. Farther downstream, the April to early June) and once in spring (mid August river receives perennial groundwater input from to late September), except for a few ephemeral springs (e.g. sites 11 and 13) and the beds of deep sites sampled in late summer after fl ooding. While permanent pools (e.g. sites 10 and 12) as it passes this regime was chosen in order to sample both through the south-western tip of Millstream the post-wet season (autumn) and the dry season National Park, after which the lower Fortescue is (spring), a dry summer in 2004/5 and a wet winter again a well-defi ned single channel with numerous in 2005 meant that rainfall patterns during the permanent pools (e.g. site 76). survey were not entirely representative of long-term Numerous relatively short rivers, such as the annual patterns. Most sites were sampled either Maitland, Sherlock and Yule, form the Port Hedland in spring and the following autumn (43 sites) or Coast Basin to the north of the Fortescue Basin. The vice-versa (13 sites) (Table 1). The rest were sampled longer of the rivers arise on the northern slopes of with more than 12 months between samples, either the Chichester Range and fl ow across the broad in spring and then autumn (23) or vice-versa (5) Roebourne Plain, while others have ill-defined or were sampled in spring and the following late headwaters on the plain’s clay flats. The lower summer (2 sites) or vice-versa (1 site), or were reaches of the Port Hedland Coast rivers contain sampled in one season only (5 in autumn, 1 in pools that are clear and spring-fed (site 56), tidally spring and 5 in late summer). infl uenced (site 18) or turbid (site 79), while in the Of the wetlands sampled (Table 1), 18 were Chichester Range the rivers contain numerous claypans and flooded clay flats, ranging from springs (site 55). Some shallow pools occur on the those that were very small (< 0.25 ha), shallow (< rocky Burrup Peninsula near Karratha (e.g. sites 75 20 cm) and ephemeral to larger, deeper examples and 77). that would hold water for many months after The large De Grey River Basin occupies most of fl ooding. Eighteen were fl owing springs and/or the eastern third of the Pilbara and includes several spring-fed creeks. Forty-one were river pools with major sub-catchments (Oakover, Shaw, Coongan clear water and coarse sediments and another and Nullagine) that form the De Grey River nine were turbid pools with clay sediments, the channel after the Oakover-Nullagine confl uence. latter mostly in smaller creeks. There were also 210 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae Spring Dates Latitude Longitude Autumn Dates sampled during the Pilbara Biological Survey. Wetlands Sitecode Site name Wetland Type Drainage Basin Summer/ PSW001PSW002 Coondiner PoolPSW003A East Marsh Fortescue (Moojari) West Marsh Fortescue PSW003B (Moojari)PSW004 West Marsh Fortescue PSW005 GnalkaPSW006 Gnoona Claypan Koondepindawarrina PoolPSW007 marsh salt Falls Fortescue PSW008 marsh salt Hamersely GorgePSW009 Bobswim PoolPSW010 marsh salt Creek Palm Spring on Cave PSW011 turbid pool Millstream at Pool Palm PSW012 Fortescue Millstream Delta turbid poolPSW013 claypan Fortescue Gregory Gorge PoolPSW014 Palm Spring Millstream at PSW015 Fortescue Creek Myanore PoolPSW016 Fortescue spring fed creek Chalyarn Pool 29/04/2006 creek fed PSW017 spring spring fed creek Nyeetbury Spring Fortescue 29/04/2006PSW018 pool clear river Pool Wodgina PSW019 clear pool river Pool Homestead Munda Ashburton 23/04/2006 17/08/2003 FortescuePSW020 Fortescue Station Yarrie at De Grey spring fed creek 23/05/2004 FortescuePSW021 clear pool Coppin spring river fed creek GapPSW022 Fortescue - 22.317 24/05/2004 Pelican Pool 16/08/2003PSW023 Ashburton turbid pool 30/08/2006 Glen Herring Pool 18/08/2003, 15/08/2003 119.149 20/05/2005PSW024 Fortescue Tanguin Rockhole 22.118 24/05/2004 21/05/2005PSW025 Fortescue Fortescue 22.508 21/05/2005 Carawine GorgePSW026 118.396 spring clear pool fed river pool/creek clear pool river Bamboo 18/08/2003, Spring 22.724 3/05/2005 119.778 30/08/2006 22/08/2003 poolPSW027 clear river 22.328 19/05/2005 Spring Wolli Weeli 19/08/2003PSW028 Onslow Coast 119.656 22.154 clear pool river 119.188 20/08/2003 Onslow Coast Desert Queen BathsPSW029 9/05/2004 Creek Pool Watrara 22.268 8/05/2004 24/08/2003PSW030 13/05/2005 118.475 Hedland Port Coast Onslow Coast 21/08/2003 Poonamerrala 22.476 Creek Pool De Grey 117.038 22.256 15/05/2005 13/05/2006 1/05/2005 Moreton 118.553 Pool spring/spring Hedland Port Coast fed creek 26/08/2003 clear river pool 117.988 21.569 25/08/2003 clear pool river 24/08/2003 23.058 clear pool river 14/05/2005 De Grey 1/05/2005 117.057 29/08/2003 118.091 pool clear river 27/08/2003 31/08/2003 21.572 29/04/2005 21.550 21.583 spring fed creek spring fed creek De Grey 28/08/2003 116.962 pool clear river De Grey clear 30/08/2003 pool river De Grey 21.858 116.971 117.069 21.442 20.522 clear pool river De Grey 21/05/2004 116.516 115.863 1/09/2003 118.060 21.753 Fortescue De Grey 21.426 River Rudall 116.038 Rudall River 118.473 16/05/2004 Rudall River 1/09/2003 28/04/2005 20.679 17/05/2004 claypan 120.201 27/04/2005 28/05/2004 22/05/2005 30/05/2004 2/09/2003 25/04/2005 20.883 2/09/2003 3/09/2003 28/05/2004 120.119 4/09/2003 8/09/2003 6/09/2003 21.342 9/09/2003 5/09/2003 21.332 21.190 8/09/2003 Ashburton 119.614 120.374 121.084 21.483 22.467 22.916 22.618 22.148 121.028 22.504 122.259 119.206 122.359 119.679 122.077 5/05/2004 28/08/2004 22.676 115.716 Table 1 Table Distribution patterns of aquatic invertebrates 211 on Ashburtonsh Pool River clear pool river Ashburton 21/04/2005 3/09/2004 23.366 116.959 PSW031PSW032 Cane Claypan River PSW033 Well Amy Near Mt. Creek Pool PSW034 Cooliarin PoolPSW035 Paradise PoolPSW036 Claypan DeGrey PSW037 Munreemya BillabongPSW038 clear pool river Coppin PoolPSW039 Sweet Well ClaypanPSW040 claypan Roy HillPSW041 Claypan Downs Mulga Outcamp Claypan Onslow CoastPSW042 Ethel Creek ClaypanPSW043 JacksonsPSW044 Bore Claypan Marsh Salt turbid Weelarrana poolPSW045 claypan 7/05/2004 Pool claypan Yandabiddy claypan claypanPSW046 Spring Fork AshburtonPSW047 Red Hill Creek Pool claypanPSW048 on Cane 25/08/2004 River Pool House PSW049 claypan Port near Henry Hedland White Pool Cliffs River CoastPSW050 claypan claypan Whiskey on Ashburton Pool RiverPSW051 6/05/2004 20/04/2006 De Grey 22.195 marsh salt on Duck Creek Grey De Pool Wallarook Fortescue turbid De Grey poolPSW052 Cheela Spring 115.955 PSW053 clear pool river Berringarra De Grey ClaypanPSW054 12/09/2004 17/08/2005 clear pool river pool clear river Catfi pool clear river PSW055 De Grey Wackilina clear pool river Creek PoolPSW056 pool clear river 26/05/2004 19/05/2004 14/05/2004 Fortescue Errawallana Spring Ashburton Ashburton Fortescue 14/05/2004 21.806 20.506 fed creek spring/spring FortescuePSW057 River Spring on Sherlock Pool AshburtonPSW058 Ashburton Ashburton 115.108 Coast Onslow 118.622 Carleecarleethong on DeGrey Pool River 20/05/2004PSW059 31/08/2006 19/09/2004 Onslow Coast Kumina Creek 13/09/2004 AshburtonPSW060 20/05/2004 The Springs River on Jones clear river pool 3/06/2004 - 18/05/2005PSW061 1/06/2004 16/05/2005 Chinaman Spring claypan 23/05/2004 22.369PSW062 2/06/2004 18/05/2005 20.670 spring fed creek pool clear river 4/06/2004 5/06/2004 16/05/2005 Miningarra 20.376 CreekPSW063 18/09/2004 turbid 118.945 pool - 120.226 Gregory Range Creek 29/08/2004PSW064 18/05/2006 119.423 27/08/2005 27/08/2004 Skull Springs De fed creekGrey spring/spring PSW065 30/08/2004 26/08/2005 20.295 - 26/08/2004 Bonnie Pool 27/08/2005 5/09/2004 Hedland Port Ashburton Coast 5/09/2004PSW066 20.570 Hedland Coast Port fed creek spring/spring Cookes Creek 22.724 Pool 119.423 31/08/2004 22.826 22.096 14/05/2005 120.258 on Kalgan Kalgan Creek Pool Hedland Port Coast 2/05/2005 22.531 Ashburton 22.675 115.698 Ashburton 21.963 120.259 22.675 115.621 23.903 22.851 23.284 pool clear river 4/05/2005 115.284 119.811 30/04/2005 116.049 22.479 24.071 118.710 19/05/2005 118.102 10/09/2004 119.811 fed creek spring/spring 120.234 11/09/2004 ephemeral creek 116.469 119.874 ephemeral De Grey creek 17/05/2006 15/09/2004 16/05/2006 pool clear river 14/09/2004 Coast Onslow 21.630 1/09/2004 20.814 117.771 De Grey fed creek spring/spring De 117.595 Grey clear river pool 21.103 2/09/2004 20.465 9/09/2004 De Grey clear pool river 22.957 4/05/2005 29/04/2005 Fortescue 117.249 119.529 116.840 23.493 22.383 DeGrey 18/09/2004 15/09/2004 DeGrey 21/04/2006 117.199 117.613 26/04/2005 24/04/2005 - 20.612 21.863 19/09/2004 120.344 116.909 21/09/2004 24/09/2004 26/04/2005 25/04/2005 20/09/2004 20.749 21.863 23.188 23/09/2004 120.307 22/09/2004 121.008 119.699 21.306 121.148 21.681 21.978 120.481 119.991 212 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae Spring Dates Latitude Longitude Autumn Dates at at Hedland Coast Port Port Hedland Coast 3/02/2006 4/02/2006 21.064 20.876 117.709 116.730 Sitecode Site name Wetland Type Drainage Basin Summer/ PSW067PSW068 Wannagunna Spring PoolPSW069 Spring) Flat Rocks Spring (=Paperbark PSW070 Pool Horrigans PSW071 Running Waters springPSW072 Karratha Granite RockPSW073 Curara ClaypanPSW074 clear pool river PSW075 Mindaroo Claypan Yarraloola PSW076 Claypan Burrup Peninsula CreekPSW077 Fortescue Lower PoolPSW078 Ashburton Burrup Peninsula Rockhole AshburtonPSW079 Claypan Croyden pool river clear PSW080A rock pools Creek Pool Peawah West fed creek spring/spring PSW080B pools) Rock Red on Indee Station (lower De Grey pools) (upper Station Indee on Rock Red PSW081 22/04/2005 20/04/2005 rock pools turbid poolPSW082 ephemeral creek Ashburton Pool Creek Ballin Billin turbid pool rock poolsPSW083 rock pool Creek Pool Tongololo clear pool river claypanPSW084 Onslow Coast Creek Claypan Turee 8/09/2005 9/09/2005PSW085 Junction Gorge Pool Hedland Port Coast 27/04/2005 turbid poolPSW086 23/04/2005 Creek Ross Glen Ashburton HedlandPSW087 Port Coast 12/05/2006 claypan 12/05/2005 Coast Carowina Onslow Pool 23.253 Fortescue 24.301PSW088 Hedland Port Coast 31/01/2006 Innawally Pool Hedland Coast Port 28/08/2005PSW089 pool 118.154 river 118.874 clear 7/09/2005 Rocky Island Pool 31/01/2006PSW090 13/05/2006 20/08/2005 Panorama Spring 14/05/2006 Ashburton Hedland Coast Port PSW091 17/05/2005 clear pool river 25/08/2005 21.685 River on Sherlock Kangan Pool PSW092 - claypan 23.554 20/05/2006 14/05/2006 Harding Pool River Grey De 121.126 20.567 22/08/2005PSW093 Hedland Coast Port on Cockerega Creek Cangan 118.258 clear Pool pool river 19/08/2005PSW094 16/08/2005 - 20.877 116.823 Creek Homestead 1/02/2006 De GreyPSW095 24/08/2005 21/08/2005 Creek Joffre 118.586 pool clear river 20.574 17/05/2005 pool clear river PSW096 Park National Millstream in creek 21.430 Un-named 20.909 22.664 ephemeral creek 116.808 clear pool river PSW097 Ashburton pool clear river 24/05/2006 Benmore claypan 115.683 Well 23/08/2005 20.764 clear pool river 21.334 clear pool river 116.068 116.740 PSW098 Ashburton 20.877 Flat Crabhole Croydon fed creek 118.013 spring/spring Ashburton 116.154 25/05/2006 Karratha Hedland Coast Port 17/08/2005 Crabhole Flat 118.587 Fortescue 28/08/2005 De Grey clear pool river 21.039 De Grey Hedland Coast Port 19/05/2006 26/04/2006 Fortescue De Grey 117.652 24/04/2006 21/05/2006 29/08/2005 creek ephemeral 21.969 25/04/2006 21.916 13/09/2005 120.856 Hedland Port Coast 9/09/2005 claypan 2/02/2006 22/05/2006 14/09/2005 22.167 creek 114.909 ephemeral 24/05/2006 16/05/2006 clay fl Fortescue 24/04/2006 9/09/2005 21/05/2006 121.078 - 21.097 clay fl 24.083 21.696 117.624 11/09/2005 13/09/2005 118.055 6/09/2005 Fortescue 10/09/2005 24.179 118.631 11/09/2005 29/01/2006 118.030 Hedland Port Coast 21.201 23.653 21.338 22.580 23.364 21.548 119.319 3/02/2006 118.099 117.069 121.033 120.191 30/01/2006 119.343 21.580 117.104 5/09/2006 22.392 23.249 118.257 119.599 21.046 117.663 Distribution patterns of aquatic invertebrates 213 six ephemeral creeks, four rockpools (or groups of and Mg2+), Cl- (measured colorimetrically using st - 2- pools) on granite or basalt outcrops, and two salt APHA method 4500-Cl- E 21 ed.), HCO3 and CO3 marshes (three sites at Fortescue Marsh and one in (acid titration), other major ions including silica (by Weelarrana Marsh). inductively coupled emission spectroscopy), total On each sampling occasion, two invertebrate fi lterable nitrogen and phosphorus (persulphate samples were collected: one benthic sample using oxidation and colorimetric determination) and a 250 μm mesh net to sample the major habitats fi lterable nitrate/nitrite (colorimetric determination within wadeable depth (e.g. open water, sediments, after conversion of nitrate to nitrite). The detritus, submerged vegetation and streambed concentration of total dissolved solids was used of fl owing areas, where present), and a plankton as a measure of salinity, except for some highly sample using a 50 μm mesh net to sample the water turbid waters where sum of major ions was used column and submerged vegetation in the same area due to inability to remove suspended particulates. as the fi rst sample. Each sample involved sweeping Samples for nutrient analyses were passed through for a total of 50 m (usually not contiguous). The a fi lter paper of pore size 0.45 μm and frozen in benthic samples were preserved in 100% ethanol the fi eld, except for highly turbid samples, which while the plankton sample was preserved in were frozen unfi ltered and centrifuged prior to buffered formalin. All samples were sieved analysis. Chlorophyll ‘a’, ‘b’, ‘c’ and phaeophytin st and sorted in the laboratory under dissecting ‘a’ were measured (APHA method 1020, 21 ed.) microscopes and representatives of each taxon from phytoplankton retained on glass-fi bre paper seen during sorting were removed and preserved after fi ltering at least 500 ml of water. Chlorophyll in 100% ethanol. The entire content of each sample fractions were combined for analysis. was sorted. Percentage areal cover of submerged plants Conductivity and pH were measured in the fi eld was estimated for each plant community type with a handheld meter. Depth was recorded as across the area sampled for invertebrates and the the maximum depth at which invertebrates were values summed to give a total for the site. Plant sampled (usually < 1.5 m). Other than deep river community type here refers to gross differences in pools, this approximated the depth of the wetland. composition and density (e.g. ‘sparse Chara’ versus ‘dense Chara’ or mixed ‘Potamogeton/Myriophyllum’ Water samples were collected approximately at a single site). Biomass was sampled by removing 15 cm below the water surface for laboratory living material (excluding roots) from six 25 × analysis of total dissolved solids (American Public 25 cm quadrats within each submerged plant Health Association [APHA] method 2540C, 21st community and washing to remove dead and ed.), nephelometric turbidity (using a Hach Model inorganic material. Quadrats were selected by 2100N turbidimeter), colour (spectrophotometric repeatedly throwing a square metal frame into the absorbance at 465 nm in a 4 cm cell compared to plant community several meters away from the platinum-cobalt standards), alkalinity (titration), thrower. These samples were air dried in the fi eld hardness (calculated from concentrations of Ca2+

Figure 3 Relative species richness of the Pilbara aquatic invertebrate fauna by major taxonomic group. 214 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 4 Richness and proportional richness of rare taxa per sample: A, richness of singleton taxa (maximum = 6); B, richness of singleton taxa as a proportion of total richness in the sample (maximum = 13%); C, richness of taxa with ≤ 5 records (maximum = 21); D, richness of taxa with ≤ 5 records as a proportion of richness in the sample (maximum 46%). Symbol size is proportional to richness or proportion of richness and samples with no relevant taxa are represented by an ‘x’. Lines on maps are IBRA regions. Some small symbols are hidden by larger symbols representing a different season at the same site. within calico bags, dried at 100ºC for 48 h in the Organic content within the fi ne sediments was laboratory and weighed to the nearest 0.01 g. Dry determined by Metson’s colorimetric modifi cation weight of each community type was then scaled of the Walkley and Black method (Metson 1956). according to the estimated percent coverage of that Strahler stream order (Strahler 1952) and community (as above), and scaled dry weights of all distances between wetland stream channels communities were summed to estimate submerged (for lotic sites) were determined from 1:250,000 2 plant biomass per m . Percent areal cover of topographic maps. Straight-line distance from emergent plants was estimated for the whole coast and geodesic distances between sites were site. Percentage cover of emergent macrophytes calculated using GIS software. Altitude was across the area sampled for invertebrates was also determined using site localities and a digital estimated by eye. elevation model and these were used to calculate Percentage areal cover (at the bed surface) climate variables using Anuclim (Australian of three substrate classes was estimated in the National University 2000). In the absence of field (bedrock+boulder, cobble+pebble and fine hydrological data for most wetlands, wetlands sediments). At least 500 ml of sediment was were assigned subjectively one of four permanency collected from several points within the area(s) values. These were 1 = ephemeral (wetland would of fi ne sediments and this was analysed in the be dry for much longer periods than wet in most laboratory for particle size composition (gravel = 2 years), 2 = seasonal/episodic (would retain water to ≤ 32 mm, sand = 0.02 to ≤ 2 mm and silt+clay = < for extended periods after fl ooding but would dry 0.02 mm). The percentage of each of these fi ne size out in most years), 3 = near permanent (wetland classes was scaled by the estimated percent areal would retain water all year in most years but coverage of fi ne sediments in the fi eld. Sediments would dry out during prolonged dry periods), 4 = from some wetlands were sampled only in one permanent (wetland would never dry out under season if it could be safely assumed that sediment current climatic regime). Characteristics of pools composition had not changed (e.g. claypans). used to assign wetlands to a hydroperiod class Distribution patterns of aquatic invertebrates 215 included local knowledge, depth, size, setting (e.g. whole survey or estimated to occur in the region. within a gorge or exposed), presence of springs and/or barriers to hyporheic fl ow. Community composition and broadscale factors Non-metric multidimensional scaling (nMDS, Invertebrate identifi cation 50 starts) and analysis of similarity (anosim) Invertebrates from the benthic and plankton routines in Primer (Primer-E Ltd. 2008) were samples were combined prior to being identifi ed used to examine relationships between wetland to species level. Morphospecies names were community composition and a priori wetland type, assigned to undescribed species or to species stream order, season, hydroperiod class, IBRA that were represented only by immature stages subregion and drainage basin. Primer’s Bray-Curtis that could not be associated with described similarity index was used and the anosim analyses adults. The taxon code listed in Appendix 2 is a were run with 999 randomisations. Combinations phylogenetically based system maintained by the of wetland type and IBRA subregion or drainage Victorian Environmental Protection Authority. All basin with few representatives (< 4 samples) were identifi cations are consistent with a survey of the excluded to avoid the effects of outliers. Two- Wheatbelt region of WA (Pinder et al. 2004) and dimensional ordinations are shown unless 3D most (other than the dipterans) are consistent with ordinations were more informative and/or had a survey of the southern Carnarvon Basin (Halse et more acceptable stress. al. 2000). Consistency of identifi cation was achieved by the use of a voucher collection consisting of Modelling assemblage distributions specimens verifi ed or determined by specialists Several assemblages, each comprising species (see Acknowledgements) with the notable exception with similar patterns of occurrence and, of dipterans. While most cladocerans and rotifers presumably therefore, with similar responses to have been identified to species level, some are environmental gradients, were recognised from associated only tentatively with named species a matrix of species occurrences by sample, with and further work is required to resolve these species and samples ordered by hierarchical identifi cations. cluster analyses (un-weighted pair grouping using arithmetic averaging [UPGMA] using default beta Data analysis of –0.1 in PATN 3.11, Belbin 2006). Two-step and Bray-Curtis dissimilarity indices were used for Sampling effi ciency species and samples respectively. Singleton species, Three sites (a claypan, a river pool and a spring) protozoans and some taxa, such as nematodes, were sampled by taking fi ve replicate 10 m sweeps likely to represent multiple species (as indicated in rather than one 50 m sweep for both plankton and Appendix 2), were excluded from analyses. benthos in order to estimate sampling effi ciency, Assemblages comprising species with similar with each 10 m sample being taken through the environmental responses should occur at different same combination of habitats. Both regional and sites as compositionally nested subsets. That wetland sampling effi ciency was analysed using is, at a location and time where environmental the EstimateS software (Colwell 2004) to run the variables are suboptimal, the assemblage should Chao2 species richness estimator adapted from be a species-poor subset of what it is at a location (Chao 1987) and the ICE estimator (Incidence-based and time where environmental conditions are more Coverage Estimator) adapted from Lee and Chao favourable (McKenzie et al. 2004). For example, (1994). an assemblage of species primarily united by a physiological requirement for freshwater will be Diversity poorly represented in a brackish water site, but Sample richness (alpha diversity) refers to the the species that do occur will be a subset of those number of species recorded at a wetland on a occurring where water is fresher. Other species particular date. Beta diversity is the rate of species might be more likely to occur in the brackish site turnover between sampling events. Whittaker’s Bw but they would belong to another assemblage. was used as a measure of beta diversity because of When such assemblages can be identified, its simplicity, robustness to inconsistencies between richness–environment relationships provide datasets, such as alpha richness and sample size insights into the environmental requirements of (Wilson and Shmida 1984), and usefulness for the component members of the assemblage. We presence/absence data and measuring ‘overall’ assessed nestedness qualitatively, using nMDS regional beta diversity rather than changes in ordinations to determine whether composition of species turnover along a gradient (Wilson and each assemblage converges as richness increases Shmida 1984; Magurran 2004). Regional richness (not shown). Where this was not the case, the refers to the number of species collected during the assemblage was subdivided and each sub- 216 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 5 2D nMDS ordination of all samples according to species composition from a priori wetland types as indicated, with Anosim global R and signifi cance. Axis labels not shown because data rotated to maximize use of the space. assemblage tested again, sometimes with the result by zero, but if the value is less than the threshold that some or all of the original assemblage was not then the constant is multiplied by the threshold modeled. Leathwick et al.’s (2005) MARS package minus the value. This means that there is a for R (R Development Core Team 2008) was used to response to the variable only below the threshold. model assemblage richness against abiotic variables The Leathwick and Elith routine generates basis (those with an * in Appendix 1). Strongly correlated functions for each variable and builds models 2 variables (r > 0.9) were not used in the same using forward-stepwise addition and backwards modeling procedure. This method identifi es sub- pruning of these, with generalised cross-validation ranges of explanatory variables, each of which has to assess model error at each step. Independent a linear relationship (determined using generalised variables were allowed into the model-building linear models) with assemblage richness and process provided they showed an ecologically creates a basis function for each (Friedman 1991; tenable relationship to richness in univariate plots Hastie et al. 2001). Basis functions take the form that did not refl ect outliers or gaps in sampling of a coefficient multiplied by the maximum of that would lead to over-fi tting. The penalty value either zero (if the measured value of the variable is K (Hastie et al. 2001) was adjusted as required to outside the range of the variable in the function) or give more parsimonious models. Simpler models the difference between the value and a threshold were preferred where this did not substantially representing one or other end of the function’s reduce adjusted r2. Variables selected by the model- range), for example: building process whose basis functions seemed to Log(richness) = constant x (maximum of ‘0’ or refl ect gaps or outliers in the independent data or ‘threshold – value’) which had a non-signifi cant relationship to richness In this case, if the value of the variable is greater (P < 0.05), were removed from the pool of available than the threshold then the constant is multiplied variables and the model-selection process re-run. Distribution patterns of aquatic invertebrates 217

Where more than one variable was thus removed were dominated by sodium, largely with sequentially, all were individually allowed back Mg2+>Ca2+>K+ (mostly river pools and springs) or into the model-selection process (and the model Ca2+>Mg2+>K+ (mostly temporary wetlands such reassessed) before a fi nal model was produced. as claypans, shallow turbid pools in creeks and A Poisson error distribution was assumed for rockpools). Some claypans and turbid creek pools assemblage richness, and frequency plots of model had Na+>Ca2+>K+>Mg2+, Na+>K+>Ca2+>Mg2+ or residuals showed that errors did approximate Na+>Mg2+>K+>Ca2+. In some of these samples, Na+ Poisson distributions, albeit with greater right skew had only marginally higher concentration than than expected. Models were assessed by calculating Mg2+ or Ca2+. Some springs, and a few river pools, adjusted r2 values and 10-fold cross-validation to were dominated by Mg2+, either with Na+>Ca2+>K+ provide mean and standard error estimates of the or Ca2+>Na+>K+. Several sites, mostly turbid pools regression intercept and slope between predicted and claypans but also a few pools and springs, and observed richness. were dominated by Ca2+, mostly with Na+>Mg2+>K+ or Mg2+>Na+>K+. Claypans and turbid pools tended RESULTS to have higher %K+ and lower %Mg2+ compared to clear pools and springs, even where the order of dominance was not different. About half the Chemical and physical features sites (largely clear river pools and springs) were Water chemistry, sediment, macrophyte and - 2- dominated by chloride, mostly with HCO3 >SO4 climate data are provided in Appendix 1. Almost 2- - but some with SO4 >HCO3 . Remaining sites, all wetlands were fresh, with salinity > 3 g/L including almost all the claypans and turbid pools, found only in some drying river pools, tidally but also some clear river pools, were dominated influenced pools and a couple of inland salt - 2- by bicarbonate, almost all with Cl >SO4 , though a marshes and creeks. Three-quarters of sites - 2- - few turbid creek pools had HCO3 >SO4 >Cl . Two

Figure 6 Bray-Curtis dissimilarity values between all samples versus: A, geodesic distance; B, channel distance be- tween lotic wetlands in the De Grey Oakover Basin; and C, channel distances between lotic wetlands in the Ashburton Basin. 218 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 7 2D nMDS ordinations of invertebrate community composition of the four major wetland types showing au- tumn versus spring samples, with anosim global R and signifi cance values shown. western Fortescue Marsh samples from site 3 had those sampled are likely to retain water all year 2- - - SO4 >Cl >HCO3 . in most years. About a third of the river pools had Three wetlands were marginally acidic (6.5 fl owing water in one or both seasons, mostly as an to 6.9) but otherwise there was an even spread infl owing stream fed by groundwater or as fl ow of pH from neutral to 10.2. A wide range of from an upstream pool and/or water fl owing out of turbidity was measured, with many turbid river the pool. Some lower-order creeks were ephemeral. pools and claypans having turbidity > 250 NTU Small shallow claypans are assumed to have been and, exceptionally, > 10,000 (maximum 200,000). ephemeral, while larger examples would have been Suspended solids seem to have inflated colour seasonal. and total fi lterable nitrogen and phosphorus at Sediment composition varied widely, ranging some claypan sites, probably because of binding of from river pools dominated by bedrock and nutrients to clay particles and incomplete removal cobbles through to more sand-dominated pools of suspended particles through centrifugation and claypans with fi ne sediments. Within rivers, (for samples too turbid to fi lter). For this reason, the main change as stream order increased from nutrient concentrations and colour were not used in one to seven was a reduction in the proportion of modeling. bedrock/boulder and a corresponding increase There are few active fl ow-gauging stations in in cobble/pebble, gravel and sand, but not silt and the region and wetlands were visited only once clay. Sediment organic content was generally under or twice, hence knowledge of their hydrological 1%, but a small number of springs had higher regimes is sketchy. All springs had discharge, values where organic matter had built up around though this was too weak to measure where water seepage areas. was only seeping from the source. Many river Emergent macrophytes usually occupied < 10% pools are known to be permanent and most of of the sampled area but a few springs and pools Distribution patterns of aquatic invertebrates 219

Figure 8 2D nMDS ordinations of: A, river pools sampled in spring and the following autumn without summer fl oods; B, river pools sampled in spring and the following autumn with summer fl oods; C, river pools sampled in spring either before a summer fl ood or before a summer drought; D, river pools sampled in autumn after a summer fl ood or after a summer drought.

(and one claypan) had emergents across > 50% (M.N. Lyons pers. comm.). Biomass of submerged of the area sampled. The emergent vegetation of macrophytes ranged from zero or close to it (at river and creek pools in the Pilbara was dominated about half the sites) to 341.4 g/m2 and was mostly by Schoenoplectus subulatus and Typha domingensis under 100 g/m2. (M.N. Lyons, DEC, pers. comm.). These species typically formed dense beds along banks and Invertebrates shallower backwaters. Cyperus vaginatus and Sampling effi ciency Eleocharis spp. occurred in briefl y inundated sites and the shallowly inundated edges of pools and The total number of species collected in fi ve 10 m springs. Small claypans were often devoid of sweep-net samples on one occasion from a spring emergent plants. At larger examples of claypans (site 9), a claypan (site 31) and a river pool (site 90) and crab-hole fl ats the dominant vegetation was were 143, 72 and 133, respectively. For these three usually emergent grass, commonly Eragrostis sets of samples, species richness estimators in benthamii but sometimes Leptochloa spp. and EstimateS produced curves that almost reached Pseudoraphis spinescens. Eucalyptus victrix also an asymptote by the fi fth sample. These estimates occurred across the bed and along the margins of suggested that we had collected 72% to 82% of the some claypans. total richness present within the portion of each About half of the sites had < 10% cover of wetland sampled. submerged macrophytes and about a quarter Diversity had more than 50% cover. Common taxa were Najas, Potamogeton, Myriophyllum and Characeae Altogether, 1034 discrete taxa (marked with an 220 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 9 A: Box and whisker plots of species richness versus permanence category for all samples (black bars = insect richness, white = non-insects); B, same for riverine samples only; C and D, axis 1 versus axes 2 and 3 of a 3D nMDS ordination of sample invertebrate composition, with samples coded by permanence class as per Appendix 1. For A and B the boxes are 25th and 75th percentiles with a median point, whiskers are 1.5 times standard deviation and open circles are outliers.

* in Appendix 2) were recorded during the study, 3 to 57 and averaged 25±0.8. Macroinvertebrate though a few of these were higher taxa which richness (all other taxa) ranged from 4 to 170 and undoubtedly represent multiple species. There averaged 69±2. Total richness at sites sampled twice were many specimens that could not be fully ranged from 37 to 265 and averaged 147±5. Beta identifi ed because they were the wrong sex or were diversity was 854/91 = 9.4 for the samples taken in too immature. About half the collected species are spring and 836/84 = 10.0 for the summer/autumn known to be described and some of the informally samples. A lower B of 7.2 was obtained when the named larvae may also represent species described w two sampling occasions were combined for each site. on the basis of adults. Richness per sample ranged from 14 to 226 and averaged 94±3 (s.e.). Regional richness was estimated to be at least Microinvertebrate richness (protozoans, rotifers, 1166 (ICE) to 1217 (Chao2). The estimator curves did copepods, cladocerans and ostracods) ranged from not reach an asymptote but the slope of the curve Distribution patterns of aquatic invertebrates 221 represented by the fi nal 50% of samples was very show richness of rare species as proportions of total shallow, with estimates of regional richness (and sample richness, with values ranging from 0 to 6% actual number of taxa collected) increasing only by for singletons and from 0 to 23% for those with ≤ 5 about one species per additional sample after 189 records. A small number of sites had particularly samples. high representation of rare species, but these were Insects comprised 42% of taxa collected, mostly mostly saline and/or ephemeral wetlands. Rare dipterans (17.9%) and beetles (10.8%) (Figure 3). species were no richer in these sites but made up a Other species-rich faunal groups were rotifers larger proportion of the community because total (20.9%), water mites (8.7%) and microcrustacea: richness tended to be lower in saline and ephemeral ostracods (6.7%), copepods (3.5%) and cladocerans wetlands. (7.3%). Leeches and leptophlebiid mayfl ies were At least 130 species, mostly rotifers and notably rare. The former was represented by a microcrustaceans and mostly undescribed, are single new species present at only four sites: two presently known only from the Pilbara. These are freshwater creeks in Rudall River National Park, an herein referred to as Pilbara ‘endemics’, though intermittent stream in Karijini National Park (Joffre we recognise that some are likely to be found Creek) and a permanent pool on the Harding River outside the region with further survey effort. This within Millstream National Park. Leptophlebiids fi gure excludes some taxa, particularly most of the had a similar pattern of occurrence, with Thraulus acarines and dipterans, that are too poorly known sp. AV1 occurring only at four sites in or near from other regions to determine distributions, but Millstream National Park and Atalophlebia sp. AV17 those taxa may also include species restricted to the occurring only at Joffre Creek and Hamersley Pilbara. Most of the endemic taxa occurred across Gorge in Karijini National Park, although both a variety of wetland types and had apparently of these species are known from outside the disjunct but not geographically restricted Pilbara (Dean 1999). Malacostracans were also distributions. There was no difference in the rare, with just one stygal isopod (Pilbaraphreatoicus proportion of the community made up by endemic platyarthiticus), two stygal amphipods (Pilbarus millsi species between any of the major wetland types or and Chydaekata sp.) and the atyid shrimp Caridina geographic subregions. indistincta. Other significant records included A number of the endemics were collected only the fairy shrimp genus Streptocephalus, recorded in turbid wetlands (claypans and/or turbid pools), from one claypan and a granite outcrop pool. The including several rotifers, hydrobiid snails, the new only other record of this genus in WA is from the fairy shrimps (Branchinella pinderi and B. mcraeae), Carnarvon Basin although it is widely but sparsely new clam shrimps (Limnadia sp. nov. and Eocyzicus distributed in inland and northern Australia (B.V. sp. nov.) and ostracods (new Paralimnocythere, Timms, University of Newcastle, pers. comm.). Limnocythere, Bennelongia and Cypretta). Several A new genus of dytiscid beetle was collected at other taxa that occurred only once during the two river pools in the central and eastern Pilbara, survey may also be claypan specialists, including one new genus of alonine cladoceran occurred two new Lynceus clam shrimps, a new genus in a lower Fortescue River pool and another was of chydorid Cladocera and several additional collected from a large claypan in the central Pilbara. ostracods. There was a tendency for these turbid Many other taxa were new to science (e.g. Pinder water specialists to occur only in wetlands of the 2008; Timms 2008; Watts and McRae 2010). western or eastern Roebourne Plains or the middle About 21% of taxa were recorded in only one to upper Fortescue Valley. sample and another 9% in just two samples. The Numerous members of this endemic component singleton and doubleton rates were about the same (mostly representing all the occurrences of when the seasonal samples were combined for particular genera or families) were restricted to a sites. On average, each taxon occurred in about 9% subset of springs and spring-fed pools. This subset of samples and in 13% of sites. Figures 4A and 4C includes wetlands in Millstream National Park show the richness of rare species in two classes: and/or in the gorges of Karijini National Park, singletons and those that occurred in fi ve or fewer as well as Running Waters and Skull Springs on samples. Rare species were distributed patchily the Oakover, Weeli Wolli Spring in the Fortescue across the Pilbara and did not exhibit a strong system and Nyeetbury Spring on the Robe River. geographic pattern. Rotifers, cladocerans and large The taxa included the undescribed leech, a branchiopods were disproportionately represented phreodrilid oligochaete (Insulodrilus sp. WA35), the in the list of singletons at the expense of most only amphipods collected (Pilbarus millsi, Chydaekata other groups, whereas taxa with ≤ 5 records were sp.), all records of isopods (Pilbarophreatoicus), taxonomically more representative of the whole most riffle beetles (Coxelmis and Austrolimnius), fauna, except that the large branchiopods were still the damselfly Nososticta pilbara and one of the represented disproportionately. Figures 4B and 4D pyralid lepidopterans (sp. 37). The same springs 222 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 10 Separate 2D nMDS ordinations of invertebrate community composition of four major wetland types with samples coded by IBRA sub-region, with anosim global R and signifi cance values shown. also supported a number of non-endemic taxa new Mytilocypris ostracod and Coxiella snails). that were similarly rare, including Wundacaenis and leptophlebiid mayflies, assiminaeid snails, Invertebrates and a priori wetland classifi cation Ecnomina and Notalina caddisfl ies and Momoniella A priori wetland types (see Introduction) were and Austraturus water mites. The hydraeinid superimposed on an nMDS ordination of all beetle Tympanogaster sp., also known only from samples based on their invertebrate community Karijini waterfalls in the Pilbara, was collected composition in Figure 5. Most clear river pools, subsequent to this survey (Jackson et al. 2009). Some springs and ephemeral creeks are distinguishable of these taxa are present in springs because they from turbid creek pools and claypans. Clear river are essentially groundwater taxa, including the pools and springs also showed relatively little isopods, amphipods and darwinulid and candonid overlap. Rockpools and ephemeral creeks both ostracods. Others belong to groups that have have variable community composition, with the strong associations with permanently moist and/ former tending to align with turbid water sites and or fl owing habitats elsewhere (the mayfl ies, leeches, the latter with clear water sites. Communities of phreodrilid oligochaetes, elmid and Tympanogaster claypans and turbid pools showed a strong overlap. beetles). Additional endemic species associated Saline river pools, such as sites 18, 29 and 53, had with other springs and river pools were more community composition more like those recorded widespread, or tended not to occur together as at some salt marshes (the four clear river pool frequently as the taxa listed above. samples near salt marshes in Figure 5). Several potentially endemic species were Overall, there was greater heterogeneity in the collected only from Fortescue Marsh (a variant fauna of claypans than riverine wetlands. An of the rotifer Brachionus angularis, a new Alona anosim analysis suggested there were signifi cant cladoceran and a new Ainudrilus oligochaete) or differences in community composition between from this wetland plus Weelarrana Salt Marsh (a most wetland types (global R = 0.494, P < 0.001), Distribution patterns of aquatic invertebrates 223

Figure 11 Separate 2D nMDS ordinations of invertebrate community composition of the four major wetland types with samples coded by drainage basin – anosim global R and signifi cance values shown. except between claypans and other turbid or similarity in community composition may be ephemeral wetlands (turbid pools, salt marshes, due to ecological processes such as dispersal, ephemeral creeks and rockpools). Since some of rather than species’ shared ecophysiological the major wetland types tended to have different responses to their environment. An example community composition, and because wetland in dryland rivers is the potential influence of types are not evenly represented across the region, recent and historical hydrological connectivity on comparisons of community composition between community composition of floodplain wetlands season, drainage basin and IBRA subregions were (e.g. Marshall et al. 2006), which may confound performed separately for each major wetland type relationships between the biota and site-based (claypans, turbid pools, clear pools and springs). variables. However, there was little evidence of spatial autocorrelation within our biological Environmental correlates of community matrix. Geodesic distance between wetlands composition was very slightly, albeit signifi cantly, correlated with Bray-Curtis community dissimilarity (r2 Spatial autocorrelation = 0.0038, P < 0.001) (Figure 6) but a Mantel test Spatial autocorrelation between biological (using the RELATE routine in Primer) showed communities (the tendency for communities in no significant rank correlation (Spearman p) close proximity to be more similar to each other between the community and geodesic dissimilarity than to more distant communities) can lead matrices (Rho = -0.048, P = 0.96). There was a to an overestimation of relationships between very low, but again signifi cant, linear correlation composition and environmental variables. between channel distance (for riverine sites) and Sampling units are not independent from Bray-Curtis community dissimilarity for the De one another (Legendre 1993) and some of the Grey-Oakover Basin (r2 = 0.0158, P < 0.05) but not 224 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 12 A and B, axis 1 versus axes 2 and 3 of a 3D nMDS ordination of all riverine samples with stream order in- dicated by numbers (stress = 0.16); C, average (± s.e.) Bray-Curtis similarity of pairs of samples within the seven stream orders, numbers of samples shown above points. for the Ashburton Basin (Figure 6). Mantel tests was most evident for some of the river pools. showed no rank correlation between community Also, the few samples in summer 2006, mostly dissimilarity and channel distance matrices for from ephemeral wetlands, tended to cluster apart either river (Rho = 0.097, P = 0.885 and Rho = 0.018, from other samples. Changed salinity or turbidity P = 0.38, respectively). These analyses suggest between seasons was probably the cause of some that relationships between faunal composition samples from the same site clustering apart, and measured environmental variables were not e.g. composition at site 53 changed signifi cantly confounded by geographic proximity to any great following an increase in salinity from 6.8 to 14 g/L. extent. Average species richness for sites was the same in autumn (85±4) as it was in spring (92±4) and none Seasonality of the individual assemblages identifi ed below was In the cluster analysis of samples (not shown), signifi cantly richer in one season than the other two-thirds of the sites sampled in both spring and (t-tests, P > 0.05). Thirty percent of species were autumn fell within the same group and frequently collected in only one season, with about the same occurred as sister samples. This indicates limited number of species absent in autumn (140) as in seasonal change in invertebrate community spring (152). However, two-thirds of these occurred composition for many sites. Occurrence of spring only once during the survey, so their occurrence and autumn samples in different cluster groups in a single season probably represents rarity rather Distribution patterns of aquatic invertebrates 225 seasonality. Average frequency of occurrence insects and there was little difference in richness within a season was the same for autumn (9.5 between wetlands classed as near permanent and samples/species) and spring (9.7) so there was permanent (ANOVA, P < 0.001 for all samples, P no tendency for species to be more common in < 0.05 for riverine samples). A 3D nMDS analysis one season. Ordinations showed little obvious (Figure 9C) showed a gradient in community separation of spring from autumn communities for composition between permanence classes 1 to 4. the main types of wetlands (Figure 7), and anosim However, this effect is diffi cult to separate from analyses suggested there was only a very slight wetland type (Figure 5) and related variables seasonal effect on community composition for such as fl ow and turbidity, because most of the clear river pools (R = 0.096) and no effect for other ephemeral sites are claypans and most of the wetland types. permanent sites are springs and river pools.

Cyclones Geography Twelve river pools were sampled in the spring Anosim and nMDS analyses (Figure 10) prior to cyclonic summer flooding (see Figure suggested that there were no differences in 2) and again in the following autumn (in 2003/4 community composition between IBRA subregions or 2005/6), while thirteen pools were sampled for the four major wetland types. In an initial in spring and the following autumn without analysis, clear river pools appeared to differ intervening fl oods (2004/5). Neither season nor between some subregions but this refl ected the occurrence of flooding had a significant effect infl uence of three saline pools (almost all other on species richness in these pools (F = 0.005 for pools were fresh to sub-saline). Once these sites season, 0.016 for fl oods, 1.1819 for interaction, P were removed from the analysis there were no > 0.1 for all terms). Figure 8 shows the results of significant differences between clear river pool four separate nMDS ordinations of these river communities of different subregions. Ordinations pool communities by season and presence/ suggested differences in community composition absence of fl ooding. Figures 8a and 8b suggest between some drainage basins, especially for there was a greater difference between spring springs and claypans (Figure 11). Anosim analyses and autumn communities where there was an confirmed that there were some differences intervening cyclone and that post-fl ood autumn between communities of springs of the Fortescue communities were more heterogeneous than Basin and those of both the Ashburton and De Grey autumn communities that did not experience basins (R = 0.39 and 0.41 respectively) and between fl ooding, although the group of sites experiencing communities of claypans in the Port Hedland Coast a cyclone was more variable beforehand (Figure Basin and those of both the De Grey and Ashburton 8c). Anosim analyses showed weak significant basins (R = 0.62 and 0.39 respectively). differences between seasonal communities whether An nMDS ordination of riverine sites indicated or not floods occurred, though the effect was some separation of samples from 1st and 2nd slightly larger where there was a fl ood (R = 0.24, P order streams from those collected in higher < 0.01 for sites with fl oods and R = 0.19, P < 0.01 for order streams (Figure 12A,B), but an anosim sites without fl ooding). Figures 8c and 8d indicate analysis suggested that differences in community greater separation of the two groups of pools composition between stream orders were negligible from each other in autumn (after fl ooding) than in (Global R = 0.091, P < 0.001). There were signifi cant spring (before fl ooding). Anosim analyses indicated differences between 1st and 2nd order streams weak, albeit signifi cant, separation of communities and between these lower orders and some of the that experienced fl ooding from those that did not intermediate orders (1st and 4th, 1st and 5th and 2nd in both seasons but differences were greater in and 5th) but signifi cant pairwise R values were autumn (R = 0.10, P < 0.01 in spring, R = 0.22, P < small (< 0.22). There was greater community 0.01 in autumn). heterogeneity among the lower order samples than among samples from higher order sites (Figure Hydroperiod 12C). The latter refl ects greater physical diversity Both richness and species composition showed among low order sites, which included both differences between estimated permanence classes. headwater springs with clear fl owing water and Figure 9 shows richness of insects (capable of rocky substrates and static creek pools on clay fl ats aerial colonisation) and other invertebrates of with turbid water and fi ne sediments. High order the four permanence classes across all samples sites tended to be more uniform, with clear water (Figure 9A) and in riverine samples only (Figure and sandy to cobbly beds. 9B). Both insect and non-insect richness increased as permanence increased from ephemeral to near- Assemblage distributions permanent, though the increase was greater for Figure 13 shows the results of the cluster analysis 226 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae of species. At the lowest level of similarity (< 0.45) richness and the MARS models. there are three groups. Group 1 includes species that tended to occur primarily in freshwater Assemblage 1 wetlands with longer hydroperiods. Within this This contains 153 species from a wide range group, seven subgroups (assemblages 1 to 7) of taxonomic groups. Higher richness was most differed primarily in their occurrence in relation strongly associated with increasing alkalinity (up to to hydroperiod, lentic versus lotic wetlands, fl ow 315 mg/L), biomass of submerged macrophytes (up and turbidity. The second group, assemblage 8 to 80 g/m2) and depth (up to 77 cm), but richness below, comprises a small number of species that declined once salinity exceeded about 3 g/L and mostly co-occurred in pools and springs in or near as %K+ increased to 13% (Figure 14). The positive Millstream National Park and, to a less extent, a few association between richness and median annual other springs. The third major group of species was temperature range, up to about 29°C, refl ects the those with a tendency to occur in more ephemeral, slightly higher richness in wetlands in the central more highly turbid or saline lentic wetlands. This band of the Pilbara. The model incorporating group included several subgroups (assemblages 9 to these variables had an adjusted r2 of 0.67. The 13) that differed in their occurrence primarily along assemblage was also associated with medium- those gradients. grained sediments (high gravel, low silt, some These assemblages, and their distribution along bedrock) and with the presence of fl owing water environmental gradients, are described in more (at least at one end of a river pool). Overall, this detail below. Where possible, MARS models assemblage is best represented in longer-lasting were developed to investigate which combination springs and river pools with abundant macrophytes and ranges of variables best explain distribution and gravelly sediments. The assemblage was patterns. Table 2 shows the number of species under-represented in the Fortescue and Roebourne in each assemblage, the median and maximum Plains subregions, which have few springs, but was representation in samples and r2 values for the otherwise geographically widespread (Figure 14) MARS models. Most assemblages tended to and represented in almost all samples. become compositionally more similar as richness increased, except for 12 and 13. Assemblages 8, 9, Assemblage 2 12 and 13 were not modelled because of their small This assemblage is the sister group to the size, uniformly low representation and/or lack of previous assemblage and comprised 52 species, nested structure (Table 2). Figures 14 to 19 show most of which were common, widespread insects maps with richness of each assemblage in autumn and the rest microcrustaceans and rotifers. Its and spring, graphs showing some of the univariate widespread, generalist nature is refl ected in its relationships between environmental variables and

Table 2 Summary statistics for assemblages and performance of MARS models.

Assemblage Number of taxa Median Average Maximum Composition MARS model in assemblage richness per richness per richness per converges adjusted r2 sample sample sample as richness increases

11533436.1102strongly0.67 2521817.637strongly0.45 3 42 1 1.6 31 moderately 0.34 4 97 4 8.1 43 strongly 0.73 56944.918strongly0.28 6 56 5 7.8 30 moderately 0.6 714733.112moderately0.25 82400.414weakly - 92100.59weakly - 10 39 0 1.4 12 moderately 0.63 11 40 0 0.7 10 weakly 0.62 12 20 0 0.4 9 no - 13 10 0 0.2 10 no - Distribution patterns of aquatic invertebrates 227

Figure 13 Results of a cluster analysis (UPGMA) of species according to their presence/absence in all 189 samples, showing assemblages 1 to 13 with summaries of their patterns of occurrence. Vertical lines are Bray-Curtis similarity values. relatively weak relationships with the measured of small, near-permanent to permanent, clear river environmental variables. The MARS model had an pools, particularly site 21 (a seasonal river pool with adjusted r2 of just 0.45 and indicated that richness 31 species) and/or site 16 (a small spring-fed pool was negatively correlated with salinity (above 3.3 with 15 species) in spring 2003. The assemblage g/L) and water temperature, but increased with also occurred in a range of other wetlands at low depth (up to 77 cm), silica (to 53 mg/L) and % sand densities (Figure 15). The sample collected from site (above 16%). Other associations not included in the 21 in spring 2003 was removed before modelling model include higher richness in seasonal to near- because of its strong leverage. Higher richness was permanent wetlands and wetlands with higher weakly associated (MARS adjusted r2 = 0.34) with macrophyte cover. Assemblage 2 was thus best increasing alkalinity (to 345 mg/L), bicarbonate represented in clear seasonal to near permanent (to 23 mg/L), submerged macrophyte biomass (to fresh river pools with sandy sediments (Figure 14). 78 g/m2) and lower annual mean temperatures (below 26°C). The latter, and an association between Assemblage 3 high richness and higher altitude, reflected a These 42 species, almost all insects, rotifers and tendency for the assemblage to occur primarily on water mites, appear to have clustered together the Chichester and northern Hamersley Ranges. largely because of their co-occurrence at a number There were also positive associations with longer 228 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae hydroperiod, low turbidity, high %Mg2+ and associated with increasing %K+ (to 7%) and % sand gravelly sediments. (to 30%) but declining %Mg2+ (to 30%) and fl ow (to 0.56 m/s) (Figure 16) in a model with adjusted r2 of Assemblage 4 0.6. These were most frequently recorded in turbid This assemblage of 97 species, over two-thirds of claypans and creek pools, most of which were in which were insects, occurred in 147 samples from the Fortescue and Roebourne subregions, and in 86 wetlands. Several members of this assemblage Fortescue Marsh when salinity was low (Figure 16). have known associations with flowing water Assemblage 7 (e.g. Austrolimnius sp. 2 and Simulium clathrinum) or with hyporheic or groundwater habitats (e.g. A large group of 148 uncommon species, about Pilbarophreatoicus platyarthricus and Darwinula 30% rotifers and 40% insects present in 166 samples stevensoni). Richness increased with fl ow (to 0.35 from 97 wetlands, but reaching a maximum of m/s), higher cover of emergent macrophytes, higher only 12 species in a sample. Composition did %Mg2+ (above about 13%) and longer hydroperiod, not converge strongly with increasing richness, but declined at temperatures exceeding 18°C. The suggesting that it is a loose grouping of species, MARS model incorporating these variables had an probably with varying environmental responses adjusted r2 value of 0.73 (Figure 15). The radiation not amenable to quantitative analysis of ecological seasonality and annual rainfall parameters in tolerances. This is reflected in the poor model the model refl ect highest richness in the central performance (adjusted r2 = 0.25). Higher richness northern Pilbara. These species were most likely was weakly associated with longer hydroperiod, to be encountered in fl owing vegetated springs, increasing flow (to 0.35 m/s), depth (to 91 cm) but also occurred in spring and hyporheic fed and alkalinity (to 87 mg/L but also with higher river pools, particularly in the Hamersley and altitude, to 237 m) and with latitudes north of 22°S. eastern Chichester subregions, plus the Millstream The latter two variables refl ect higher richness in wetlands (western Chichester; Figure 15). the Chichester and northern Hamersley Ranges (Figure 17). Richer sites also tended to have coarser Assemblage 5 sediments and have higher %Mg2+. This patchily This assemblage comprised 69 species, half of distributed assemblage showed little fidelity to which were microinvertebrates (rotifers, ostracods any particular wetland type, though was generally and cladocerans) and the rest almost all insects, richer in longer-lasting springs and clear river pools present in 173 samples from 97 wetlands. Richness with fl ow than in claypans and turbid river pools. was highest in seasonal to near-permanent Assemblage 8 freshwater wetlands with low fl ow, low alkalinity, intermediate %Mg2+ (richness peaking around 16%), A small group of 24 species that clustered moderate depth (richness increasing to about 75 together largely because they tended to co-occur cm) and low to moderate turbidity (Figure 16). The in a number of Fortescue River wetlands in or near MARS model incorporated some of these variables Millstream National Park (Figure 17), with sporadic but had an adjusted r2 of only 0.28 and tended to occurrences elsewhere. Millstream species included grossly underestimate high richness. These species the short-range endemic damselfly Nososticta occurred in a wide range of wetlands, but were best pilbara, the dragonfl y Nanophlebia injibandi, one of represented in mildly turbid claypans and static the two leptophlebiid mayfl ies (Thraulus sp.), two river pools rather than in springs and highly turbid of the four species of elmid beetles, the caenid claypans. There was also a tendency for richness mayfly Wundacaenis dostini and one of the two to be higher in the northern and eastern Pilbara, amphipods recorded during the survey (Pilbarus especially in autumn (Figure 16). millsi). A few members of this assemblage, such as Chydaekata amphipods and a Gondwanabates water Assemblage 6 mite, also occurred at one or more other permanent This was a small group of 56 taxa, about half spring-fed sites (especially Running Waters, Weeli insects and the rest mostly microcrustaceans Wolli Spring and Skull Springs) in the southern and rotifers. Several species, such as Bennelongia and eastern Pilbara. This appears to be a group of ostracods and the conchostracan Caenestheriella species with a preference for permanent fl owing packardi commonly occur in turbid ephemeral water. Due to the small size and very sparse waters. The species occurring in Fortescue Marsh distribution of this assemblage, its richness was not and some other saline wetlands, such as the modelled. dragonfl y Diplacodes bipunctata and rotifer Polyarthra Assemblage 9 dolichoptera, are freshwater species tolerant of low salinity rather than halophiles (cf. some members A group of 21 species, largely microcrustacea of assemblage 13). High richness was most closely and rotifers, collected in just 41 samples from Distribution patterns of aquatic invertebrates 229

Figure 14 Results of MARS modeling and maps showing richness of assemblages 1 and 2 in spring and autumn. Shown for each assemblage are graphs of richness versus selected environmental variables, the model equation, ad- justed r2 and the slope and intercept of the regression line with standard deviations. See Appendix 1 for units of environmental variables. Sizes of symbols on the maps are proportional to maximum richness in either season. Internal boundaries on the maps are IBRA sub-regions. 230 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 15 Results of MARS modeling and maps showing richness of assemblages 3 and 4 in spring and autumn. Shown for each assemblage are graphs of richness versus selected environmental variables, the model equation, adjusted r2 and slope and intercept of the regression line with standard deviations. See Appendix 1 for units of environmental variables. Sizes of symbols on the maps are proportional to maximum richness in either season. Internal boundaries on the maps are IBRA sub-regions. Distribution patterns of aquatic invertebrates 231

Figure 16 Results of MARS modeling and maps showing richness of assemblages 5 and 6 in spring and autumn. Shown for each assemblage are graphs of richness versus selected environmental variables, the model equation, adjusted r2 and slope and intercept of the regression line with standard deviations. See Appendix 1 for units of environmental variables. Sizes of symbols on the maps are proportional to maximum richness in either season. Internal boundaries on the maps are IBRA sub-regions. 232 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 17 Assemblages 7 to 9, showing (for all three assemblages) maps of richness in spring and autumn and (for as- semblage 7 only) graphs of richness versus selected environmental variables, the model equation, adjusted r2 and slope and intercept of the regression line with standard deviations. See Appendix 1 for units of envi- ronmental variables. Sizes of symbols on the maps are scaled to the maximum richness recorded in either season. Internal boundaries of the maps are IBRA sub-regions. Distribution patterns of aquatic invertebrates 233

Figure 18 Results of MARS modeling and maps showing richness of assemblages 10 and 11 in spring and autumn. Shown for each assemblage are graphs of richness versus selected environmental variables, the model equa- tion, adjusted r2 and slope and intercept of the regression line with standard deviations. See Appendix 1 for units of environmental variables. Sizes of symbols on the maps are proportional to maximum richness in either season. Internal boundaries on the maps are IBRA sub-regions. 234 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Figure 19 Assemblages 12 and 13, showing (for both assemblages) maps of richness in spring and autumn and (for as- semblage 12) univariate relationships with assemblage richness. See Appendix 1 for units of environmental variables. Sizes of symbols on the maps are scaled to the maximum richness recorded in either season. Inter- nal boundaries on the maps are IBRA sub-regions.

30 wetlands, most of which were mildly turbid 419 (Carnarvon Basin) and the copepod Calamoecia claypans, rockpools, turbid river pools and some baylyi. Other species are new, including the new other non-fl owing river pools. These were collected Eocyzicus conchostracan, which is probably also a mostly from wetlands on the coastal plains, lower claypan specialist. As with assemblage 11 below, De Grey-Oakover catchment and middle to upper some species (e.g. the beetle Berosus nutans and Fortescue Valley, where most of the claypans occur hemipteran Anisops gratus) are likely to be present (Figure 17). Only four wetlands had more than three simply because the wetlands are lentic rather representatives and the assemblage was not modeled. than because of a preference for ephemerality or turbidity. The distribution of this assemblage Assemblage 10 reflected the distribution of claypans—mostly A group comprised of 39 species, about half coastal and along the middle Fortescue Valley microcrustacea and rotifers and the rest mostly (Figure 18), as indicated by the bimodal response beetles and dipterans, present in 81 samples from of richness to altitude (Roebourne Plains with the 55 wetlands. Included are several species known lowest altitude and middle to upper Fortescue to be associated with turbid waters, including the Valley at about 400 m Figure 18). Richness was ostracods Ilyocypris sp. nov. (also recorded from the highest at ephemeral to seasonal wetlands with northern WA wheatbelt), and Limnocytheridae sp. moderate turbidities (richness increasing between Distribution patterns of aquatic invertebrates 235 about 11 and 440 NTU) and in wetlands with low well represented in the survey but there are some Mg2+ and Cl- (refl ecting the ionic composition of gaps where further sampling is likely to add to claypans). The MARS model, incorporating these knowledge of the region’s aquatic invertebrate variables, had an adjusted r2 of 0.63. These showed diversity. Almost all sampling was undertaken a strong preference for mildly to moderately turbid in either autumn or spring, but communities in waters. ephemeral wetlands sampled in late summer tended to cluster apart from spring and autumn samples. A Assemblage 11 few species (e.g. Aedes pseudonormanensis, Calamoecia Only 41 samples from 31 wetlands contained any ultima and Limnadia sp. P1) were more common in of these 41 species. Included are species known to these few summer samples than in the many spring be claypan or ephemeral wetland specialists, such or autumn samples, so there may be an immediate as Branchinella and Streptocephalus fairy shrimps, post-fl ood fauna in ephemeral wetlands that was the cladoceran Australospilus elongatus, shield under-sampled by this survey. Signifi cant areas that shrimps (Triops), the copepod Calamoecia halsei and were poorly sampled, or not sampled, include the Bennelongia ostracods. Several new species are wetlands in dunes south of the Kunderong Range, probably also claypan specialists, including new south-west of Newman, which were not sampled Lecane and Keratella rotifers, Cypretta ostracods and due to diffi cult access. Areas of mulga fl ats fed by Branchinella fairy shrimp. The model suggests that endorheic drainage in the Hamersley Range (e.g. high richness was most strongly associated with Mount Bruce, Wanna Munna and Coondawanna short hydroperiods, increasing turbidity (above 10 Flats) were dry whenever visited during the survey NTU), declining alkalinity (below 155 mg/L), low and also represent a wetland type not sampled. gravel content in sediments (below 8%) and low High-order river pools above tidal infl uence on the submerged macrophyte cover (below 30%). The Roebourne Plains were under-sampled. Finally, the assemblage is thus most prevalent in ephemeral middle Fortescue River between Millstream and and highly turbid claypans, but with some species Mulga Downs station, which is not so well defi ned also present in ephemeral creeks and gnammas on as most other major Pilbara rivers, and which rock outcrops. includes some braided and deltaic sections, was not sampled. Assemblage 12 These gaps notwithstanding, the sampling Twenty species, primarily microcrustaceans, regime appears to have captured a very signifi cant rotifers and dipterans, occurring in 29 samples proportion of the aquatic invertebrate diversity from 21 wetlands (Figure 19). These are largely inhabiting the region’s dominant wetland types. species already known to be halotolerant or There is no consensus on what proportion of site or halophilic, such as Hexarthra spp. and Tanytarsus regional richness is required to have been sampled barbitarsus, but some, including Coxiella sp. and before conclusions about biodiversity patterning Mytilocypris sp. nov., are probably additional can be drawn, other than that it should be high halophiles. These tended to occur in the more (New 1998). Chao et al. (2009) suggested that 90–95% saline wetlands with higher pH, fi ne sediments, may be a reasonable objective, but did not comment low biomass of submerged macrophytes (but not on how adequate this might be for various low cover) and low turbidity. However, a MARS purposes. Our analyses suggest that we have model that converged could not be found. captured up to 85% of the regional fauna inhabiting Assemblage 13 the types of wetlands sampled, though this is probably an overestimate for the region considering Ten species, mostly microcrustaceans and rotifers, the lack of an asymptote on the richness estimator seemingly united by their co-occurrence in one curves and the sampling gaps identifi ed above. sample (Yarraloola Claypan, site 74, in autumn, Recent collecting in Pilbara river pools (Pinder Figure 19), but very sporadically occurring in a few and Leung 2009) revealed an additional 18 other wetlands (18 samples from 19 wetlands). Some macroinvertebrate taxa from 10 high-order river are commonly found in ephemeral habitats (e.g. the pools on the Roebourne Plains. Given that the work rotifer Lecane curvicornis, the cladoceran Macrothrix excluded rotifers and microcrustacea, this number fl abelligera and beetle Sternolophus immariginatus). of additional species is greater than predicted from This assemblage was not modelled due to its sparse richness estimator curves, but lowland river pools and probably artifi cial nature. are one of the sampling gaps noted above. The proportion of the estimated regional fauna DISCUSSION that we collected in the Pilbara survey appears to be higher than for most other large-scale biodiversity Adequacy of sampling surveys. In a survey of aquatic invertebrates in The common wetland types of the Pilbara were the Carnarvon Basin, Halse et al. (2000) recorded 236 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

476 species, which is 78% of the regional richness the limited previous sampling effort means that estimated using the same methods. However, the most of the species in listed in Appendix 2 are new estimate of regional richness did not quite reach records for the region and many are new records an asymptote and the observed and estimated for Western Australia. The 10% of species that are regional richness were still increasing by one ‘new to science’ is similar to results of surveys of or two species per sample near the end of the other Western Australian regions (e.g. Halse et al. curve. In the wheatbelt region of south-western 2000; Pinder et al. 2004). Australia, Pinder et al. (2004) collected 957 species, With an estimate of about 1200 species of aquatic which is about 73% of the estimated regional invertebrates, and an average of 94 species recorded richness, but again the curve did not fl atten out per sample, the Pilbara fauna appears to be rich. and observed and estimated regional richness This is in contrast to some other studies that were increasing by about two taxa per sample near have suggested that Australian arid zones have the end of the curve. Compared to these studies, a depauperate fauna (e.g. Marshall et al. 2006), and probably most others listed below, we appear but few of these studies have had the taxonomic to have sampled a comparatively high proportion breadth and depth of the Pilbara survey. The of the fauna in the Pilbara. Undetected taxa are Carnarvon Basin, south-west of the Pilbara, has undoubtedly rare at the regional scale and are similar annual rainfall (200–400 mm) but is only unlikely to have altered outcomes of our analyses a third the size of the Pilbara. Halse et al. (2000), signifi cantly (Arscott et al. 2006). using the same sampling protocol and taxonomic At the individual wetland scale, the sampling scope and resolution as used in the Pilbara, but methods employed captured about 75–85% of with half the number of samples and sites, recorded species estimated to have been present at three 476 micro and macroinvertebrate taxa, from at least test sites. These are very similar to estimates of 600 estimated to be present in the region. Average invertebrate sampling efficiency recorded by richness was only 31 taxa per sample and 41 per site Halse et al. (2002), who concluded that this was (two dates combined for most sites). Lower regional suffi cient to contrast invertebrate faunas of different richness would be expected in a smaller, more wetlands. However, for some larger heterogeneous topographically homogeneous region, especially wetlands, such as Carawine Gorge and Paradise since a greater proportion of the wetlands sampled Pool, use of a single sampling site is unlikely by Halse et al. (2000) were ephemeral and/or saline. to have captured such a large proportion of the However, even for the same wetland types, alpha wetlands’ overall diversity. richness seemed to be lower in the Carnarvon Basin. Richness in river pools (38±2) and claypans Diversity (24±3) was only about a third of that recorded for While there are some taxonomic publications such wetlands in the Pilbara. Salinity in these that focus on north-west faunas, such as Lansbury wetland types was similar when sampled in both (1984) and Smit (1997, 1998), and many others projects, but groundwater is mostly saline in the with some Pilbara records, there are very few Carnarvon Basin and higher salinities may, at studies providing information about the aquatic times, represent a bottle-neck for invertebrate invertebrate communities of the Pilbara region. diversity. Carnarvon Basin river pools also tend As part of a continental-scale river condition to have fi ner and more homogeneous sediments assessment, Kay et al. (1999, supplemented by compared to the Pilbara (unpublished data, see later data) recorded 76 macroinvertebrate families Halse et al. 2006) which reduces habitat diversity. from 18 river pools, including four families not Whatever the cause, lower richness in river pools recorded in this survey. Ponder (1987) undertook represents a smaller species pool available to a small survey of springs (Millstream, Fortescue colonise the more ephemeral wetlands during Falls, Running Waters and Skull Springs) and fl oods or via aerial dispersal (Anderson and Smith concluded that these contained rare and/or 2004). restricted species, which was confi rmed during The semi-arid wheatbelt region is comparable in this study. Halse et al. (2002) recorded 159 species size to the Pilbara but has little permanent water (including 18 groundwater species) in a study that (and virtually no permanent freshwater other than examined the extent to which springs provide reservoirs), despite higher rainfall (300–600 mm habitat for both groundwater and surface water per annum) and signifi cant interaction between faunas. Our survey recorded most of these species surface water and the mostly saline groundwater. but few additional groundwater taxa (detailed A survey of the wheatbelt (Pinder et al. 2004), using below). Pilbara odonates were fairly well known protocols and sampling effort similar to those prior to this survey (Watson 1969) and our survey used in the Pilbara, collected 957 taxa from an supports the established view that Nososticta pilbara estimate of > 1300 in the region, but alpha diversity is restricted to the Millstream wetlands. However, was lower (average 40±2) and species turnover Distribution patterns of aquatic invertebrates 237 or beta diversity much greater (βw = 24) than in of taxa, in the Pilbara we collected 6 to 190 per the Pilbara. The high proportion of ephemeral wetland with an average of 93. and/or naturally or secondarily saline wetlands There are few comparable datasets from arid included in the survey refl ects their preponderance zones outside Australia. Hall et al. (2004) collected in the wheatbelt but contributed to the low alpha 107 macroinvertebrate taxa, mostly identifi ed to diversity. The very high diversity of wetland species level, from 73 freshwater playas in Texas, types, large latitudinal range of the study area and each sampled three times. Richness at a wetland numerous salt lakes with strongly heterogeneous ranged from 24 to 33 per sampling event, compared faunas probably explain the high beta and regional with 10 to 98 (median 55) for equivalent wetlands diversity in the wheatbelt. in the Pilbara (claypans), though some of the The arid Lake Eyre Basin (LEB) is larger than Pilbara claypans were ten times as large as the the Pilbara (approximately 1.1 million km2) but has largest Texan playas. De Moor et al. (2004) collected lower annual rainfall (average 125 mm). This basin 216 macroinvertebrate species from 27 riverine has very few permanent fresh waterbodies (Bunn et sites along a 340 km stretch of the permanently al. 2006), some waterholes are saline and most are fl owing Cunene River in south-western Africa. This reliant on surface fl ows rather than groundwater, compares to 325 to 398 macroinvertebrate species so salinity often rises between fl ood events due to in fewer pools along the longest Pilbara rivers evapoconcentration (Costelloe 2004) and/or saline (Ashburton, Fortescue, De Grey-Oakover). seepage. In studies of this region’s aquatic fauna, Based on these limited comparisons with other Shiel et al. (2006) collected 423 zooplankton species parts of Australia and other continents, the Pilbara (protozoans, rotifers, copepods and cladocerans) appears to have particularly high diversity at the from 56 wetlands in the LEB. This is comparable wetland level, although it is less outstanding at with the 346 species in these faunal groups the regional level because of low beta diversity. A collected in the Pilbara, given that protozoans were partial explanation of the observed richness is the not well sampled in the Pilbara survey. Madden et high sampling effort, broad taxonomic scope and al. (2002), Costelloe et al. (2004) and Marshall et al. fi ne taxonomic resolution employed in this survey, (2006) collected about 170 macroinvertebrate taxa but aspects of Pilbara wetlands may also promote (mostly identifi ed to genus) from about 90 wetlands high diversity. Invertebrate richness in desert rivers in the southern and eastern LEB. The number of has been associated with the duration of fl ow and macroinvertebrate species is likely to be at least water permanence, substrate diversity, habitat twice the number of genera (macroinvertebrate complexity and proximity to permanent refugia species to genus ratios were 2 to 2.5 in the (Sheldon et al. 2002; Boulton et al. 2006). Pilbara Pilbara and other large Western Australian river pools tend to have high habitat diversity, with surveys mentioned above), so the regional aquatic multiple emergent and submerged macrophyte invertebrate fauna is likely to be over 800 species, communities, organic litter and debris, fl owing especially since ostracods were not identified. water at one or both ends of the pool (often from Data on richness at individual wetlands was not springs or alluvial discharge), dense submerged provided in these studies. root mats along pool edges and a variety of Most wetlands in the arid Paroo region of inland inorganic substrates (often with clay, gravel, cobble New South Wales are intermittent, although and bedrock in the same pool). This does not many fi ll in most years. Timms and Boulton (2001) explain high diversity in Pilbara claypans, however, recorded 217 invertebrate taxa in 87 wetlands which are as structurally simple as those elsewhere, in this region. Some groups (e.g. ostracods and although high biotic diversity in riverine refugia cladocerans) were under-sampled or not identifi ed may promote high fl oodplain diversity following to species, while rotifers were rarely identified fl oods. at all, so it is likely the species list substantially Although the Pilbara is an arid zone, its geology underestimates richness of individual wetlands (extensive aquifers) and climate (relatively regular as well as the region. Richness of individual summer rainfall occurring in intense events) wetlands ranged from 10 to 36 in saline lakes, 39 result in an abundance of permanent fresh refugia, to 71 in freshwater lakes and 27 to 29 in claypans. including springs and river pools maintained by Compared to the present survey, samples were springs and hyporheic fl ows. Permanent waters much smaller but were collected more frequently. affect regional and wetland richness in at least In central Australia, Davis (1997), Davis et two ways. Firstly, the abundance of permanent al. (1993) and Brim Box et al. (2008) recorded water, combined with the latitudinal position of 130 macroinvertebrate species from 42 central the Pilbara, is likely to have allowed both southern Australian wetlands. Richness ranged from 4 to 64 and tropical taxa to extend into the region in a per wetland but several groups likely to be speciose way that may be uncommon in arid regions (see were not identifi ed to species. For the same range Biogeography below). Secondly, in arid zones 238 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae permanent waters tend to have higher species and a few extending around both northern and richness than temporary waters (Halse et al. 1998; western Australia. About 5% have broader Western Brendonck and Williams 2000; Sheldon et al. 2002; Australian distributions, occurring primarily south Eitam et al. 2004; Boulton et al. 2006; Brim Box of the Pilbara, including a few apparently arid et al. 2008; Waterkeyn et al. 2008). In our survey, zone species not known outside the state. About invertebrate richness of wetlands increased with 2% have broad southern Australian distributions increasing hydroperiod class, especially amongst and are at their known north-western limit in the more mobile insects which are able to colonise the Pilbara. The regionally endemic component river pools after they have become isolated. Longer described earlier comprises the remaining 19%. hydroperiod allows more species to colonise and It is very likely that some of these will be found greater habitat complexity to develop (Boulton and outside the Pilbara with further survey effort. For Jenkins 1998; Sheldon et al. 2002). Longer-lasting instance, the newly described Haliplus fortescuensis wetlands are also frequently deeper and larger, Watts and McRae, 2010, first collected during which may also promote greater habitat diversity. the present project, was recently recorded in the An abundance of permanent wetlands may also Kimberley region (Daniel et al. 2009). Similarly, translate to high regional richness by allowing numerous species fi rst collected in the Carnarvon the region to maintain a large species pool during Basin (Halse et al. 2000) have now been collected drought periods and by increasing the probability in the Pilbara (e.g. Keratella sp. nov., Bennelongia of colonisation by irregular immigrants. Regular sp. 414, Branchinella halsei). The Pilbara fauna is frequency of inundation also promotes high thus a blend of widespread species, species near invertebrate diversity in non-permanent wetlands their northern limits, species shared with tropical (Boulton and Lloyd 1992) and, in the Pilbara, few northern Australia that rarely extend further south, wetlands would go more than a couple of years plus a north-west fauna. Absent from the Pilbara is without some inundation. a larger tropical fauna that does not extend west of the Kimberley region. This includes many genera In the river pools of some arid zones, water and species of odonates, mayfl ies, leeches, and most quality deteriorates as water volume declines species of atyid shrimps and palaemonid prawns during drought periods, even in permanent (though one palaemonid, Macrobrachium, has wetlands, and especially where water levels are recently been collected from the lower De Grey by controlled by surface water fl ows and/or where Pinder and Leung [2009]). Short et al. (2004) suggest saline groundwater intrudes when water levels that north Queensland may have been a centre are low (Costelloe et al. 2005; Hamilton et al. for dispersal of species of Macrobrachium to other 2005). Such pools can experience raised salinity parts of tropical Australia owing to its proximity to and temperature, concentration of nutrients and Papua New Guinea (where Macrobrachium diversity increased solar radiation as pools contract away is higher), and that regional diversity in Australia from banks, with subsequent problems associated may be related to distance from north Queensland with algal blooms (Lake 2003). These impacts (the Kimberley region is relatively depauperate). have well-known effects on aquatic invertebrates, They also suggest that past and present aridity including reduced species richness (reviewed in may limit occurrence of Macrobrachium outside of Lake 2003). Hamilton et al. (2005) suggested that the tropics and, in this respect, the Great Sandy waterholes maintained by fresh groundwater Desert is probably a formidable barrier to westward would be ‘exceptionally important’ refugia in the extension of many tropical species. Some species Cooper Creek system of the Lake Eyre Basin. By restricted to the Pilbara have congeners that are contrast, in the Pilbara, pools maintained by fresh restricted to either the Kimberley or to the northern hyporheic or spring fl ows are not unusual and tropics more broadly (e.g. the damselfl ies Eurysticta water quality tends to remain high during dry coolawanyah and Nososticta pilbara in the Pilbara and periods. This probably also contributes to high Eurysticta kununurra and Nososticta kalumburu in invertebrate richness between fl oods. the Kimberley region) and the Pilbara species may represent lineage divergence through isolation. Biogeography The limited subregional patterning within the There is suffi cient information for about two- Pilbara was associated with the uneven geographic thirds of the species collected to comment on the distribution of wetland habitats. Assemblages fauna’s biogeographic affi nities. About half of the occupying springs and spring-fed pools were species are widespread in Australia or their known richer in upland areas (where there is more surface distributions are patchy. These do not have well- groundwater discharge) whereas those occupying defi ned, sub-continental distributions, though a turbid waters were richer in the valleys and few occur mostly in the inland. About a quarter lowlands. The faunas of springs and claypans also have a northern distribution, with a few of these showed some differentiation between drainage also occurring in Asia, some extending well inland basins. Species in these habitats are generally Distribution patterns of aquatic invertebrates 239 less mobile and less likely to be moved around the assumption that arid wetlands are episodic by fl oods, possibly contributing to subregionally (infrequently and unpredictably fl ooded) rather distinctive faunas. River pool faunas were more than intermittent (more regularly and predictably uniform across the region and showed little flooded). Regional endemism in intermittent longitudinal variation along rivers. The lack of wetlands is actually very common (Simovich strong patterning in rivers within the region is 1998; Watts 1999; Halse et al. 2000; Costelloe et al. consistent with results of a survey of north-western 2004; Pinder et al. 2004; Shiel et al. 2006), whereas fi sh faunas (Morgan and Gill 2004) which showed episodic wetlands seem to have much less regional that, while the Pilbara has a distinct fi sh fauna, endemism, at least in Australia (Timms 2001, 2007). compared to the west-fl owing rivers south of the Temporary wetlands in the Pilbara are generally Pilbara, it represents a single biodiversity province intermittent rather than episodic, so the presence for fi sh. of a regionally endemic component to their faunas Davis (1997) and Brim Box et al. (2008) suggested would be expected. that permanent waters of central Australia are Despite 19% of the Pilbara fauna possibly being important because they provide a refuge from regionally endemic, our data suggest that very few drought for indigenous species and mesic relicts. of the region’s surface water invertebrates are likely In the Pilbara this is true of the subset of springs to be short-range endemics (sensu Harvey 2002). and spring-fed pools that support a small mesic- While many species were encountered rarely, their adapted fauna (including assemblage 8), some occurrences tended to be widely separated. Rare of which are probably regional endemics. These exceptions to this include the Pilbara damselfl y sites include the Millstream wetlands, springs in Nososticta pilbara (restricted to Millstream wetlands) Karijini gorges and a few other springs. A few of and the fairy shrimp Branchinella pinderi (known these species are groundwater taxa, for which the only from the western Roebourne Plains). springs are potential points of dispersal rather The stygofauna component of the Pilbara than refuges, but most belong to faunal groups survey has revealed a groundwater fauna that is that have known associations with permanently particularly diverse compared to similarly sized moist habitats. It is not clear why this group of regions elsewhere, with 350 species recorded from species is seemingly restricted to only a subset of an estimate of at least 500 species for the region the springs, although differences in the hydrology (Eberhard et al. 2005, 2009; S.A. Halse, pers. obs.). and vegetation structure of the springs may be part Since many springs and river pools are areas of the explanation. The springs supporting these of contact between groundwater and surface species are shaded by tall riparian woodlands or water, it might be expected that such wetlands are within deep gorges, have strong base fl ows, would be zones of overlap between stygal and and most have some deep pools and lacked a dense epigean faunas. We recorded most of the 18 stygal cover of benthic algae. By contrast, springs without species that Halse et al. (2002) recorded in Pilbara this faunal component tended to lack canopy springs, plus a few additional ones: a number of cover and/or had just trickles or seepage when phreodrilid oligochaetes, a copepod (Parastenocaris sampled, or had just a shallow creek rather than jane), two ostracods (Humphreyscandona adorea deep pools, and frequently had substrates covered and Sarscypridopsis cf. ochracea), plus the with dense fi lamentous algae. Another explanation amphipods Chydaekata and Pilbarus and the isopod may be that the few springs supporting this mesic Pilbarophreatoicus platyarthricus. These occurred element are those that have been consistently wet mostly in springs but some were recorded in for longer periods. The occurrence of mesic species permanent river pools with spring inflows or restricted to the most permanent wetlands in coarse sediments that presumably had hyporheic central Australia led Davis (1997) and Brim Box et fl ow. However, the occurrence of stygal species in al. (2008) to suggest that conservation of permanent surface water wetlands was relatively uncommon wetlands could protect more native biological compared to the frequency with which epigean diversity than conservation of temporary ones. To species were recorded in groundwater (S.A. Halse, some extent this is borne out in the Pilbara, where pers. comm.). a quarter of the regionally endemic component of the fauna was collected only from permanent river Environmental correlates of biodiversity patterns pools and springs, compared to 12% collected only The 13 species assemblages clearly differed in in ephemeral to seasonal wetlands. However, for their associations with water chemistry, habitat this 12%, and for many other species, conservation variables, geographic distribution and climate: of temporary wetlands is critical. some responded to the same variables but with Williams (1998) and Williams and Kokkin (1988) different slopes, thresholds or polarity. This argued that arid zone wetlands tend not to support provides some understanding of how the region’s regionally endemic species, but this was based on aquatic invertebrate fauna is distributed across 240 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae many of the major environmental gradients. these for our study sites. A number of major fl ood The predictive power of the models produced events during the survey did allow us to investigate for nine of these assemblages was rarely high, the effects of these on seasonal changes in but nonetheless the models highlighted which invertebrate communities. Numerous studies have combinations of variables were most strongly demonstrated the resilience to major fl ood events correlated with high representation of the of the aquatic invertebrate faunas of arid zones assemblages and identifi ed response thresholds. (Fischer et al. 1982; Stanley et al. 1994; Anderson Four smaller assemblages were not amenable and Smith 2004). Our data suggest that, within a to modeling, but some of these were narrowly few months of fl ooding, river pool communities distributed in relation to environmental gradients were as rich as they were before fl ooding and (e.g. halophilic species in assemblage 12 and species as rich as pools that did not experience fl oods. of permanently fl owing springs in assemblage 8). There was some indication that pre- and post- Broad-scale surveys such this one often reveal fl ood communities had different composition but a large number of relationships between the biota the difference was not large. Thomaz et al. (2007) and the environment, not all of them causal. suggested that fl oods tend to homogenise river Climate variables showed subtle univariate faunas. Some features of riverine communities in relationships with all assemblages but few were the Pilbara are consistent with this. Thus, stream order had little bearing on overall community selected in models. Those included in models had nd primarily coastal to inland gradients (cooler and composition below 2 order, and riverine wetlands less variable temperatures and higher rainfall generally had more homogeneous communities near the coast) aligned with patterns in richness compared to fl oodplain wetlands, despite the latter generally being structurally simpler wetlands. of three assemblages (1, 3 and 4). The climate parameters were amongst the strongest parameters Where there is a salinity gradient, it is also in those models, but it is not clear whether such normally one of the strongest influences on correlations refl ect ecophysiological tolerances (e.g. aquatic invertebrates (Hammer 1986). Many to temperature or reliability of inundation) or are wetlands of arid zones are normally saline, and correlations with some other factors, such as the some freshwater wetlands become saline either distribution of particular wetland habitats not as through concentration of salts or intrusion of well described by latitude, longitude and altitude. saline groundwater as water levels decline (Timms and Boulton 2001; Costelloe et al. 2005; Shiel et al. Variables describing hydrology, salinity, turbidity, 2006). Almost all wetlands in the Pilbara were aquatic vegetation and sediment composition were fresh when surveyed, other than the two salt also associated with the distributions of different marshes, a couple of river pools influenced by assemblages. These variables are considered to tidal fl ows, and pools and streams on the Burrup be primary drivers of invertebrate community Peninsula possibly affected by sea spray. With composition in dryland rivers (Boulton and Lloyd many Pilbara river pools being maintained by 1991; Timms 2001; Timms and Boulton 2001; fresh groundwater, salinisation between flows Sheldon et al. 2002; Costelloe et al. 2004; Shiel et al. appears to be relatively uncommon in the region 2006) but their effects are diffi cult to separate in (sites 29 and 53 were exceptions in this study). survey datasets such as this. The only assemblage dominated by halotolerant There was a gradient in overall community or halophilic species was assemblage 12, which composition from ephemeral wetlands through had lowest richness in freshwater and maximum to those considered to be permanent. This is richness in the two salt marshes. Other halotolerant in line with most previous studies that have species were distributed across several assemblages examined the influence of hydroperiod and that had greatly reduced richness as salinity invertebrate communities (e.g. Boulton and Lloyd increased. Where salinity was a parameter in 1992; Collinson et al. 1995; Schneider and Frost 1996; models, it indicated a decline in richness above Eitam et al. 2004; Waterkeyn et al. 2008). Permanence about 3.3 g/L. This threshold is close to the value class and/or fl ow were associated with richness of of 4.1 g/L below which Pinder et al. (2005) found most assemblages and were parameters in some aquatic invertebrate richness to decline (2.6 when of the models. Depth may also have reflected halophilic species were excluded). Alkalinity hydroperiod: the depth thresholds in richness was also a feature in many of the models, with models were in the range 70 to 100 cm, coinciding richness responding either positively or negatively with the transition from wetlands that would to this variable. Alkalinity has been correlated have been ephemeral to those deemed to be near with invertebrate composition and biomass in permanent. A range of other hydrological variables numerous studies (Koetsier et al. 1996; Death describing timing, frequency and extent of fl ows and Joy 2004), for reasons that are not fully and inundation also infl uence wetland faunas of understood but often attributed to an effect on arid zone wetlands (Walker et al. 1995; Boulton primary production (e.g. Dickman 1973), microbial 1999) but we did not have information on most of processing of detritus or biomineralisation (Mackie Distribution patterns of aquatic invertebrates 241 and Flippance 1983). Other components of ionic may be incidental to these other factors. Where composition have physiological consequences there was a positive relationship (e.g. assemblages for aquatic invertebrates, although more so in 1, 3 and 4) there was generally an initial increase in saline waters (e.g. Radke et al. 2003). We observed richness as cover and/or biomass increased to some strong relationships between richness of many of moderate level, then no further response or even a the assemblages and proportional contributions decline in richness (e.g. response of assemblage 4 to of ions, most notably Mg2+ and K+. There may submerged macrophyte cover). This may have been have been a degree of direct biological response because habitat heterogeneity would be increased in these relationships but it is likely that these by the presence of moderate amounts macrophytes correlations were acting as surrogates for a range but reduced by more complete coverage or very of characteristics that distinguish lentic from lotic high density of macrophytes. Other studies have wetlands not represented as well by other variables. reported similar non-linear relationships (e.g. Lentic wetlands in the Pilbara had relatively high Collier et al. 1999). Strong positive or negative %K+, perhaps due to the tendency for clay soils relationships were observed between richness and to concentrate potassium more than the coarser percentage contribution of sediment size fractions, sediments of the larger river channels, whereas lotic but this relationship was usually only for the wetlands had proportionally higher %Mg2+. range before the fraction became dominant. This Turbidity is often naturally very high in arid zone might also refl ect a heterogeneity effect. In our wetlands with clay sediments (Bunn et al. 2006) survey, the single large invertebrate samples were and is certainly associated with very distinctive frequently taken across multiple sediment types communities (Timms and Boulton 2001; Timms and macrophyte communities (including absence 2002). The role that turbidity plays, independent of macrophytes) and this may have weakened our of shallow depth and ephemerality (also typical ability to describe richness-habitat relationships. of claypans), is not well understood, but it limits macrophyte growth and affects algal production CONCLUSIONS through light limitation, resulting in benthic primary productivity being largely restricted The Pilbara has a particularly high diversity of to the edge of the waterbody (Bunn et al. 2003; aquatic invertebrates for an arid zone, at least at Fellows et al. 2007). Turbidity probably also allows the level of the individual wetland. We suggest that the occurrence of the larger crustaceans such this high richness partly refl ects the abundance of as Branchinella and Triops that would otherwise consistently fresh, permanent water maintained by be prone to predation by waterbirds, although extensive groundwater aquifers. About a fi fth of the ephemerality frequently eliminates fish which fauna is presently known only from the Pilbara but would also predate on such invertebrates. Many there is also a substantial pan-northern element and of the pools in smaller lowland creeks were turbid smaller western and southern elements. The fauna in the Pilbara, as were all the claypans, with the is generally widespread within the region, with shallowest of the latter having extremely high little evidence of local endemism. Of particular note turbidity (> 10,000 NTU). This variable was present is a group of relatively rare and mostly mesic taxa in the models of the two assemblages primarily that are heterogeneously distributed across a limited occurring in claypans (10 and 11). Richness of both range of springs and spring-fed pools, including of these assemblages was associated positively those of Millstream and Karijini National Parks. with high turbidity, with assemblage 10 increasing The limited sub-regional patterning largely in richness only between 10 and 440 NTU (then refl ects the distribution of wetland habitats, such declining) and assemblage 11 increasing in richness as more ephemeral turbid waters along the coast more continuously from about 10 NTU. and broader inland valleys and more permanent The importance of macrophytes and sediments flowing water in upland areas. There were as structural features of arid zone wetlands has some differences in composition of springs and not been as much of a focus as have the roles of claypans between drainage basins but river pool hydrology, salinity and turbidity, but some authors communities are generally homogeneous within have indicated their importance (Boulton and and between basins. Aquatic invertebrate faunas of Lloyd 1991; Sheldon et al. 2002). They are certainly river pools regain similar richness and composition known to influence invertebrate communities within a few months of catastrophic fl oods. more generally (Williams and Mundie 1978; The fauna is distributed along a range of Cyr and Downing 1988; Wollheim and Lovvorn environmental gradients, including hydroperiod, 1996). Our analyses showed both positive and flow, salinity, turbidity, sediment composition negative relationships between assemblage richness and extent of aquatic vegetation. Conservation and macrophyte variables, though the negative of aquatic invertebrate diversity in the Pilbara is relationships were for species preferring turbid or dependent on the persistence of wetland habitats saline conditions, so low macrophyte abundance along these gradients. 242 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

ACKNOWLEDGEMENTS sediments of the River Murray. Regulated Rivers Research and Management 7: 137–151. Numerous pastoralists and indigenous comm- Boulton, A.J., Sheldon, F. and Jenkins, K.M. (2006). unities generously shared their knowledge of Natural disturbance and aquatic invertebrates in Pilbara wetlands (and provided the occasional desert rivers (pp. 133–153). In: Kingsford, R. (ed.), shower and bed). We particularly thank Pilbara Ecology of desert rivers. Cambridge University Press: DEC staff, including Peter Kendrick and Stephen Cambridge, U.K. van Leeuwen, for providing advice on wetland Brandis, A.J. (2009). Rescuing the rangelands : management locations, contacts and working in the Pilbara, and strategies for restoration and conservation of the natural other Pilbara staff for assisting with logistics. Leah heritage of the Western Australian rangelands after 150 Stratford assisted in the laboratory and Lauryne years of pastoralism. Western Australian Department Grant and Harley Barron assisted in the field. of Conservation and Land Management: Perth, Anna Leung and Ross Gordon helped prepare Australia. data and tables. Numerous taxonomists verifi ed or Brendonck, L. and Williams, W.D. (2000). Biodiversity in wetlands of dry regions (drylands) (pp. 181–194). In: identifi ed voucher specimens: Brian Timms (large Gopal, B., Junk, W.J. and Davis, J.A. (eds), Biodiversity branchiopods), Chris Watts (beetles), Alena Glaister in wetlands: assessment, function and conservation. Vol. 1. (Elmidae), Fredric Govedich (leeches), Phil Suter Backhuys Publishers: Leiden, Netherlands. (mayfl ies), Tom Weir (hemipterans), Ros St Clair Brim Box, J., Duguid, A., Read, R.E., Kimber, R.G., (leptocerid caddisfl ies), John Dean (hydropyschid Knapton, A., Davis, J. and Bowland, A.E. (2008). caddisflies), Hendrik Segers (rotifers), Gunther Central Australian waterbodies: the importance of Theischinger (odonates) and Winston Ponder permanence in a desert landscape. Journal of Arid (molluscs). Some climatic data were provided by Environments 72: 1395–1413. the Bureau of Meteorology’s National Climate Brooks, T.M., da Fonseca, G.A.B. and Rodrigues, A.S.L. Centre and was dated 2008. The GIS section of DEC (2004). Protected areas and species. Conservation provided geodesic distances between wetlands. Biology 18: 616–618. Bunn, S.E., Davies, P.M. and Winning, M. (2003). Sources REFERENCES of organic carbon supporting the food web of an arid zone fl oodplain river. Freshwater Biology 48: 619–635. Anderson, J. and Smith, L. (2004). Persistence and Bunn, S.E., Thoms, M., Hamilton, S.K. and Capon, S.J. colonization strategies of playa wetland invertebrates. (2006). Flow variability in dryland rivers: boom, bust Hydrobiologia 513: 77–86. and the bits in between. River Research and Applications Arscott, D.B., Jackson, J.K. and Kratzer, E.B. (2006). Role 22: 179–186. of rarity and taxonomic resolution in a regional and Bureau of Meteorology Australia (2008). Mean monthly spatial analysis of stream macroinvertebrates. Journal and mean annual rainfall and temperature data (base of the North American Benthological Society 25: 977–997. climatological datasets). Bureau of Meteorology Australian Government (2004). Interim Biogeographic Australia: Melbourne, Australia. Regionalisation of Australia v6.1. Commonwealth of Chao, A. (1987). Estimating the population size for Australia. http://www.environment.gov.au/parks/ capture-recapture data with unequal catchability. nrs/science/bioregion-framework/ibra/index.html. Biometrics 43: 783–791. Australian National University (2000). ANUCLIM 5.1. The Chao, A., Colwell, R.K., Lin, C. and Gotelli, N.J. (2009). Fenner School of Environment and Society, Australian Suffi cient sampling for asymptotic minimum species National University: Canberra, Australia. richness estimators. Ecology 90: 1125–1133. Boulter, C.A. (1986). One billion years of Archean history, Collier, K., Champion, P. and Croker, G. (1999). Pilbara Craton, Western Australia. Geology Today 106: Patch- and reach-scale dynamics of a macrophyte- 106–111. invertebrate system in a New Zealand lowland Boulton, A.J. (1999). Why variable fl ows are needed for stream. Hydrobiologia 392: 89–97. invertebrates of semi-arid rivers (pp. 113–128). In: Collinson, N., Biggs, J., Corfield, A., Hodson, M., Kingsford, R.T. (ed.), A free-fl owing river: the ecology of Walker, D., Whitfi eld, M. and Williams, P. (1995). the Paroo River. New South Wales Parks and Wildlife Temporary and permanent ponds: an assessment of Service: Sydney, Australia. the effects of drying out on the conservation value Boulton, A.J. and Jenkins, K.M. (1998). Flood regimes of aquatic macroinvertebrate communities. Biological and invertebrate communities in fl oodplain wetlands Conservation 74: 125–133. (pp. 137–148). In: Williams, W.D. (ed.), Wetlands in a Colwell, R.K. (2004). EstimateS: Statistical estimation of dry land: understanding for management. Environment species richness and shared species from samples, Australia: Canberra, Australia. Version 7. User’s Guide and application published at: Boulton, A.J. and Lloyd, L.N. (1991). Macroinvertebrate http://purl.oclc.org/estimates. assemblages in fl oodplain habitats of the lower River Costelloe, J.F. (2004). Hydrological results (pp. 175–259). Murray, South Australia. Regulated Rivers Research and In: Costelloe, J.F., Hudson, P.J., Pritchard, J.C., Management 6: 183–201. Puckridge, J.T. and Reid, J.R.W. (eds). ARIDFLO Boulton, A.J. and Lloyd, L.N. (1992). Flooding frequency Scientifi c Report: Environmental flow requirements of and invertebrate emergence from dry floodplain arid zone rivers with particular reference to the Lake Distribution patterns of aquatic invertebrates 243

Eyre drainage basin. Final report to South Australian ecosystem following flash flooding. Ecological Department of Water, Land and Biodiversity Conservation Monographs 52: 93–110. and Commonwealth Department of Environment and Friedman, J.H. (1991). Multivariate adaptive regression Heritage. School of Earth and Environmental Sciences, splines. The Annals of Statistics 19: 1–141. University of Adelaide: Adelaide, Australia. Gaston, K.J. (2000). Global patterns in biodiversity. Nature Costelloe, J.F., Grayson, R.B., McMahon, T.A. and Argent, 405: 220–227. R.M. (2005). Spatial and temporal variability of water Geological Survey of Western Australia (1990). Geology salinity in an ephemeral, arid-zone river, central and mineral resources of Western Australia. Memoir 3. Australia. Hydrology Processes 19: 3147–3166. Western Australian Geological Survey: Perth, Australia. Costelloe, J.F., Hudson, P., Pritchard, J.C., Puckridge, J.T. Haila, Y. and Margules, C.R. (1996). Survey research in and Reid, J.R.W. (2004). ARIDFLO Scientifi c Report: conservation biology. Ecography 19: 323–331. Environmental fl ow requirements of arid zone rivers with Hall, D.L., Willig, M.R., Moorhead, D.L., Sites, R.W., particular reference to the Lake Eyre drainage basin. Final Fish, E.B. and Molhagen, T.R. (2004). Aquatic report to South Australian Department of Water, Land and macroinvertebrate diversity of playa wetlands: Biodiversity Conservation and Commonwealth Department the role of landscape and island biogeographic of Environment and Heritage. School of Earth and characteristics. Wetlands 24: 77–91. Environmental Sciences, University of Adelaide: Halse, S., Shiel, R.J. and Williams, W.D. (1998). Aquatic Adelaide, Australia. invertebrates of Lake Gregory, northwestern Cyr, H. and Downing, J.A. (1988). The abundance of Australia, in relation to salinity and ionic composition. phytophilous invertebrates on different species Hydrobiologia 381: 15–29. of submerged macrophytes. Freshwater Biology 20: Halse, S.A., Cale, D., Jasinska, E.J. and Shiel, R.J. 365–374. (2002). Monitoring change in aquatic invertebrate Daniel, G., Kern, S., Pinder, A. and Nowicki, A. biodiversity: sample size, faunal elements and (2009). Resource condition report for a significant analytical methods. Aquatic Ecology 36: 395–410. Western Australian wetland: Lake Eda. Prepared for Halse, S.A., Lyons, M.N. and Pinder, A.M. (2004). the Inland Aquatic Integrity Resource Condition Biodiversity patterns and their conservation in Monitoring Project. Western Australian Department of wetlands of the Western Australian wheatbelt. Records Environment and Conservation: Perth, Australia. of the Western Australian Museum, Supplement 67: Davis, J.A. (1997). Conservation of aquatic invertebrate 337–364. communities in central Australia. Memoirs of the Halse, S.A., Scanlon, M.D. and Cocking, J.S. (2002). Do Museum of Victoria 56: 491–503. springs provide a window to the groundwater fauna Davis, J.A., Harrington, S.A. and Friend, J.A. (1993). of the Australian arid zone? Balancing the groundwater Invertebrate communities of relict streams in the budget: proceedings on CD of an International Groundwater arid zone – the George Gill Range, central Australia. Conference, 12–17 May 2002, Darwin, Australia. Australian Journal of Marine & Freshwater Research 44: Halse, S.A., Scanlon, M.D. and Cocking, J.S. (2006). 483–505. First national assessment of river health: Western de Moor, F.C., Barber-James, H.M., Harrison, A.D. and Australian program. milestone report 5 and fi nal report. Lugo-Ortiz, C.R. (2004). The macroinvertebrates of Western Australian Department of Environment and the Cunene River from the Ruacana Falls to the river Conservation: Perth, Australia. mouth and assessment of the conservation status of Halse, S.A., Shiel, R.J., Storey, A.W., Edward, D.H.D., the river. African Journal of Aquatic Science 25: 105–122. Lansbury, I., Cale, D.J. and Harvey, M.S. (2000). Dean, J.C. (1999). Preliminary keys for the identifi cation of Aquatic invertebrates and waterbirds of wetlands Australian mayfl y nymphs of the family Leptophlebiidae. and rivers of the southern Carnarvon Basin, Western Identification guide no. 20. Cooperative Centre for Australia. Records of the Western Australian Museum, Freshwater Ecology: Albury, Australia. Supplement 61: 217–265. Death, R. and Joy, M. (2004). Invertebrate community Hamilton, S.K., Bunn, S.E., Thoms, M.C. and Marshall, structure in streams of the Manawatu-Wanganui J.C. (2005). Persistence of aquatic refugia between region, New Zealand: the roles of catchment versus fl ow pulses in a dryland river system (Cooper Creek, reach scale infl uences. Freshwater Biology 49: 982–997. Australia). Limnology & Oceanography 50: 743–754. Dickman, M. (1973). Changes in periphytic algae Hammer, U.T. (1986). Saline lake ecosystems of the world. Dr. following bicarbonate additions to a small stream. W. Junk Publishers: Dordrecht, Netherlands. Journal of the Fisheries Research Board of Canada 30: Harvey, M.S. (2002). Short-range endemism among the 1882–1884. Australian fauna: some examples from non-marine Eitam, A., Blaustein, L., Van Damme, K., Dumont, H.J. environments. Invertebrate Systematics 16: 555–570. and Martens, K. (2004). Crustacean species richness Hastie, T., Tibshirani, R. and Friedman, J.H. (2001). The in temporary pools: relationships with habitat traits. elements of statistical learning: data mining, inference and Hydrobiologia 525: 125–130. prediction. Springer: New York, U.S.A. Fellows, C.S., Wos, M.L., Pollard, P.C. and Bunn, S.E. Jackson, J., Kern, S., Pinder, A., Nowicki, A. and Daniel, (2007). Ecosystem metabolism in a dryland river G. (2009). Resource condition report for a signifi cant waterhole. Marine and Freshwater Research 58: 250–262. Western Australian wetland: Fortescue Marshes. Prepared Fischer, S.G., Gray, L.J., Grimm, N.B. and Busch, D.E. for the Inland Aquatic Integrity Resource Condition (1982). Temporal succession in a desert stream Monitoring Project. Western Australian Department of 244 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

Environment and Conservation: Perth, Australia. Bioregional summary of the biodiversity audit for Johnson, S.L. and Wright, A.H. (2001). Central Pilbara Western Australia. Western Australian Department of groundwater study. Western Australian Water and Conservation and Land Management: Perth, Australia. Rivers Commission: Perth, Australia. McKenzie, N.L., Gibson, N., Keighery, G.J. and Rolfe, Kay, W.R., Smith, M.J., Pinder, A.M., McRae, J.A., Davis, J.K. (2004). Patterns in the biodiversity of terrestrial J.A. and Halse, S.A. (1999). Patterns of distribution of environments in the Western Australian wheatbelt. macroinvertebrate families in rivers of north-western Records of the Western Australian Museum, Supplement Australia. Freshwater Biology 41: 299–316. 67: 293–336. Keighery, G.J., Gibson, N., Lyons, M.N. and Burbidge, McKenzie, N.L., van Leeuwen, S. and Pinder, A.M. (2009). A.H. (2000). Flora and vegetation of the southern Introduction to the Pilbara Biodiversity Survey, Carnarvon Basin, Western Australia. Records of the 2002–2007. Records of the Western Australian Museum, Western Australian Museum, Supplement 61: 77–154. Supplement 78: 3–89. Koetsier, P., Minshall, W. and Robinson, C.T. (1996). Metson, A.I. (1956). Methods of chemical analysis of Benthos and macroinvertebrate drift in six streams soil survey samples. New Zealand D.S.I.R. Soil Bureau differing in alkalinity. Hydrobiologia 317: 41–49. Bulletin 12: 193–204. Lake, P. (2003). Ecological effects of perturbation by Morgan, D. and Gill, H. (2004). Fish fauna in inland drought in flowing waters. Freshwater Biology 48: waters of the Pilbara (Indian Ocean) Drainage 1161–1172. Division of Western Australia – evidence for three Lansbury, I. (1984). Some Nepomorpha (Corixidae, sub-provinces. Zootaxa 636: 1–43. Notonectidae and Nepidae) (Hemiptera-Heteroptera) New, T.R. (1998). Invertebrate surveys for conservation. of north-west Australia. Transactions of the Royal Oxford University Press: Oxford, U.K. Society of South Australia 108: 35–49. Nicholls, A.O. (1989). How to make biological surveys Leathwick, J.R., Roew, D., Richardson, J., Elith, J. and go further with generalised linear models. Biological Hastie, T. (2005). Using multivariate adaptive Conservation 50: 51–75. regression splines to predict the distributions of New Pinder, A.M. (2008). Phreodrilidae (Clitellata: Annelida) Zealand’s freshwater diadromous fish. Freshwater in north-western Australia with descriptions of two Biology 50: 2034–2052. new species. Records of the Western Australian Museum Lee, S.-M. and Chao, A. (1994). Estimating population 24: 459–468. size via sample coverage for closed capture-recapture Pinder, A.M., Halse, S.A., McRae, J.M. and Shiel, R.J. models. Biometrics 50: 88–97. (2004). Aquatic invertebrate assemblages of wetlands Legendre, P. (1993). Spatial autocorrelation: trouble or and rivers in the wheatbelt region of Western new paradigm? Ecology 74: 1659–1673. Australia. Records of the Western Australian Museum 67: Lyons, M.N., Gibson, N., Keighery, G.J. and Lyons, S.D. 7–37. (2004). Wetland fl ora and vegetation of the wheatbelt Pinder, A.M., Halse, S.A., McRae, J.M. and Shiel, R.J. of southwestern Australia. Records of the Western (2005). Occurrence of aquatic invertebrates of the Australian Museum, Supplement 67: 39–89. wheatbelt region of Western Australia in relation to Mackie, G. and Flippance, L. (1983). Relationships salinity. Hydrobiologia 543: 1–24. between buffering capacity of water and the size and Pinder, A.M. and Leung, A. (2009). Conservation status calcium content of freshwater mollusks. Freshwater and habitat associations of aquatic invertebrates in Pilbara Invertebrate Biology 2: 48–55. coastal river pools. Western Australian Department of Madden, C.P., McEvoy, P.K., Taylor, D.J., Tsymbal, Environment and Conservation: Perth, Australia. V., Venus, T.A. and Goonan, P.M. (2002). Ponder, W.F. (1987). Report on a preliminary survey of Macroinvertebrates of watercourses in the Lake Eyre springs in the Pilbara region of Western Australia. Basin, South Australia. Verhandlungen Internationale Australian Museum: Sydney, Australia. Vereinigung für theoretische und angewandte Limnologie Pressey, R.L. (2004). Conservation planning and 28: 591–600. biodiversity: assembling the best data for the job. Magurran, A.E. (2004). Measuring biological diversity. Conservation Biology 18: 1677–1681. Blackwell Science Ltd: Malden, U.K. Quinlan, K., Pinder, A.M. and Cale, D. (2009). Assessment Margules, C.R., Pressey, R.L. and Williams, P.H. (2002). of the aquatic fauna conservation values of wetlands in the Representing biodiversity: data and procedures for Hutt River/Hutt Lagoon catchments. Western Australian identifying priority areas for conservation. Journal of Department of Environment and Conservation: Perth, Bioscience 27: 309–326. Australia. Marshall, J.C., Sheldon, F., Thoms, M. and Choy, S. R Development Core Team (2008). R: a language and (2006). The macroinvertebrate fauna of an Australian environment for statistical computing. R Foundation for dryland river: spatial and temporal patterns and Statistical Computing: Vienna, Austria. environmental relationships. Marine and Freshwater Radke, L.C., Juggins, S., Halse, S.A., De Deckker, P. and Research 57: 61–74. Finston, T. (2003). Chemical diversity in south-eastern McKenzie, K.G., Halse, S.A. and Gibson, N. (2000). Some Australian saline lakes II: biotic implications. Marine gaps in the reserve system of the southern Carnarvon and Freshwater Research 54: 895–912. Basin, Western Australia. Records of the Western Schneider, D. and Frost, T. (1996). Habitat duration and Australian Museum, Supplement 61: 547–567. community structure in temporary ponds. Journal of McKenzie, K.G., May, J.E. and McKenna, S. (2002). the North American Benthological Society 15: 64–86. Distribution patterns of aquatic invertebrates 245

Sheldon, F., Boulton, A.J. and Puckridge, J.T. (2002). Timms, B.V. (2009). A revision of the Australian Conservation value of variable habitat connectivity: endemic clam shrimp genus Limnadopsis Spencer aquatic invertebrate assemblages of channel and and Hall (Crustacea: Branchiopoda: Spinicaudata: fl oodplain habitats of a central Australian arid-zone Limnadiidae). Records of the Australian Museum 61: river, Cooper Creek. Biological Conservation 103: 13–31. 49–72. Shiel, R.J., Costelloe, J.F., Reid, J.R.W., Hudson, P. and Timms, B.V. and Boulton, A.J. (2001). Typology of Powling, J. (2006). Zooplankton diversity and arid-zone fl oodplain wetlands of the Paroo River assemblages in arid zone rivers of the Lake Eyre (inland Australia) and the infl uence of water regime, Basin, Australia. Marine and Freshwater Research 57: turbidity, and salinity on their aquatic invertebrate 49–61. assemblages. Archiv für Hydrobiologie 153: 1–27. Simovich, M.A. (1998). Crustacean biodiversity and Vinson, M.R. and Hawkins, C.P. (1998). Biodiversity of endemism in California’s ephemeral wetlands stream insects: variation at local, basin and regional (pp. 107–114). In: Witham, C.W., Bauder, E.T., Belk, scales. Annual Review of Entomology 43: 271–293. D., Ferren, W.R. and Ornduff, R. (eds), Ecology, Walker, K.F., Sheldon, F. and Puckridge, J.T. (1995). A conservation and management of vernal pool ecosystems perspective on dryland river ecosystems. Regulated – proceedings from a 1996 conference. California Native Rivers: Research and Management 11: 85–104. Plant Society: Sacramento, U.S.A. Walshe, T.V., Halse, S.A., McKenzie, N.L. and Gibson, Smit, H. (1997). Australian water mites of the genus N. (2004). Toward identifi cation of an effi cient set Arrenurus, with the description of twelve new of conservation recovery catchments for Western species from northern and Western Australia (Acari: Australian wheatbelt biodiversity. Records of the Hydrachnellae: Arrenuridae). Records of the Western Western Australian Museum, Supplement 67: 365–384. Australian Museum, Supplement 18: 233–261. Water and Rivers Commission (1996). Pilbara region water Smit, H. (1998). The water mite family Limnesiidae from resources review and development plan, summary report northern and Western Australia (Acari: Actinedida), 4. Western Australian Water and Rivers Commission: with descriptions of two new species. Records of the Perth, Australia. Western Australian Museum 18: 347–355. Waterkeyn, A., Grillas, P., Vanschoenwinkel, B. and Stanley, E., Buschman, D., Boulton, A., Grimm, N. and Brendonck, L. (2008). Invertebrate community Fisher, S. (1994). Invertebrate resistance and resilience patterns in Mediterranean temporary wetlands along to intermittency in a desert stream. American Midland hydroperiod and salinity gradients. Freshwater Biology Naturalist 131: 288–300. 53: 1802–1822. Stork, N.E., Samways, M.J. and Eeley, H.A.C. (1996). Watson, J.A.L. (1969). Taxonomy, ecology and Inventorying and monitoring biodiversity. Trends in zoogeography of dragonflies (Odonata) from the Ecology and Evolution 11: 39–40. north-west of W.A. Journal of Zoology 17: 65–112. Strahler, A.N. (1952). Hypsometric (area-altitude) analysis Watts, C.H.S. and McRae, J.M. (2010). The identity of of erosional topography. Bulletin of the Geological Haliplus Latrielle (Coleoptera: Haliplidae) from the Society of America 63: 1117–1142. Pilbara region of Australia including the description Strayer, D.L. (2006). Challenges for freshwater of four new species. Records of the Western Australian invertebrate conservation. Journal of the North Museum 25: 387–398. American Benthological Society 25: 271–287. Watts, R.J. (1999). Biodiversity in the Paroo River and its Taylor, G. (1994). Landscapes of Australia: their nature wetlands (pp. 13–22). In: Kingsford, R.T. (ed.), A free- and evolution (pp. 60–79). In: Hill, R.S. (ed.), History fl owing river: the ecology of the Paroo River. New South of the Australian vegetation: Cretaceous to Recent. Wales National Parks and Wildlife Service: Sydney, Cambridge University Press: Cambridge, U.K. Australia. Thomaz, S., Bini, L. and Bozelli, R.L. (2007). Floods Williams, D.D. and Mundie, J.H. (1978). Substrate size increase similarity among aquatic habitats in river- selection by stream invertebrates and the infl uence of fl oodplain systems. Hydrobiologia 579: 1–13. sand. Limnology and Oceanography 23: 1030–1033. Timms, B.V. (2001). Large freshwater lakes in arid Williams, W.D. (1998). Diversity and evolution of the Australia: a review of their limnology and threats for fauna of the dryland wetlands (pp. 167–171). In: their future. Lakes and Reservoirs 6: 183–196. McComb, A.J. and Davis, J.A. (eds), Wetlands for the Timms, B.V. (2002). Limnology of the claypans of future: INTECOL’s 5th International Wetlands Conference, The Paroo, arid-zone Australia. Verhandlungen 1996. Gleneagles Press: Adelaide, Australia. Internationale Vereinigung für theoretische und Williams, W.D. and Kokkin, M.J. (1988). The angewandte Limnologie 28: 130–133. biogeographical affi nities of the fauna in episodically Timms, B.V. (2007). The biology of the saline lakes of fi lled salt lakes: a study of Lake Eyre South, Australia. central and eastern inland of Australia: a review with Hydrobiologia 158: 227–236. special reference to their biogeographical affi nities. Wilson, M.V. and Shmida, A. (1984). Measuring beta Hydrobiologia 576: 27–37. diversity with presence-absence data. Journal of Timms, B.V. (2008). Further studies on the fairy Ecology 72: 1055–1064. shrimp genus Branchinella (Crustacea, Anostraca, Wollheim, W.M. and Lovvorn, J.R. (1996). Effects Thamnocephalidae) in Western Australia, with of macrophyte growth forms on invertebrate descriptions of new species. Records of the Western communities in saline lakes of the Wyoming High Australian Museum 24: 289–306. Plains. Hydrobiologia 323: 83–96. 246 A.M. Pinder, S.A. Halse, R.J. Shiel, J.M. McRae

APPENDIX 1 Environmental variables for the 189 sampling events during the survey. Those marked with an asterisk were included in assemblage modeling.

See CD inside the back cover, or visit: http://www.museum.wa.gov.au/research/records-supplements/attachments

APPENDIX 2 Occurrence of aquatic invertebrate species in the 189 samples collected during the survey. In the matrix, 1 = presence, blank = absence, * indicates distinct species (i.e. not just juveniles or females of a taxon potentially listed in another row). Column 2 indicates reasons for excluding rows from analysis (see Methods for details): 1) protozoans, 2) higher taxa probably comprising numerous unidentifi ed species, 3) uncertain taxonomic resolution, 4) species merged to genus or family level for analysis, or 5) incompletely identifi ed specimens (e.g. wrong sex or juvenile).

See CD inside the back cover, or visit: http://www.museum.wa.gov.au/research/records-supplements/attachments

Grey-headed Honeyeater (Lichenostomus keartlandi) forages in shrub and tree canopies throughout the region (Lochman).